Official Codebase of 3DCodeVerse
Python
Use stable rubric names without version suffixes
gaoypeng2fccbae9f6d1
diff --git a/eval/bench/_fixed_eval.py b/eval/bench/_fixed_eval.pyindex 511a393c..dd4bf8cb 100644--- a/eval/bench/_fixed_eval.py+++ b/eval/bench/_fixed_eval.py@@ -35,12 +35,12 @@ from codeverse3d.judges.base import SLICE_TRACKS, JudgeInput, judged_subset, roufrom codeverse3d.proc import read_json_or_nonefrom codeverse3d.workspace import Workspace-RUBRIC = "static_object_v1"+RUBRIC = "static_object"def rubric_for(spec: Spec | Track) -> str:"""The fixed judge's rubric for a cell (or for a battery's track): the track's ``TRACK_INFO``- row — static_object_v1 / articulated_v1 / scene_v2 / shader_v2.+ row — static_object / articulated / scene / shader.The one place the compare bench maps a track to a rubric: ``FixedEvaluator`` defaults toit, ``judge_for`` picks each cell's judge with it and ``_compare_report`` names it in the@@ -129,7 +129,7 @@ class FixedEvaluator:cell of an ``ab_plan`` battery was recorded ``judge_error: no judgment`` — measured2026-08-26 on ``seed_v1`` (both arms, three rounds of in-loop verdicts each, andno pair score). Graphics cells are judged on their frame sheet with the track's- rubric (``shader_v2``), through ``LikenessJudge`` when the spec carries reference+ rubric (``shader``), through ``LikenessJudge`` when the spec carries referencephotos so the arms see the same photos the loop saw."""from codeverse3d.judges.vlm_judge import VlmJudge, judge_for@@ -165,7 +165,7 @@ class FixedEvaluator:from codeverse3d.spatial.frame_metrics import frame_gate_from_rendersfrom codeverse3d.spatial.scene_placement import placement_gate- # the deterministic placement caps (scene_v2: floating / sunken / interpenetration /+ # the deterministic placement caps (scene: floating / sunken / interpenetration /# scale, 2026-10-14) fire in every A/B cell as they do in the loop: the gate reads# the build's census and the cell's plan.json (a one-shot arm has none — the# plan-less checks still run); None when the census has no placement tablediff --git a/eval/bench/_loop_diag.py b/eval/bench/_loop_diag.pyindex 69c151cf..95d40fff 100644--- a/eval/bench/_loop_diag.py+++ b/eval/bench/_loop_diag.py@@ -99,7 +99,7 @@ class Veto:_OBJECT_VETO_KINDS = (f"{CONNECTIVITY}/*", f"{CONTRACT}/*")#: the veto per track. Graphics has none: its ERRORs are image-level verdicts (black / blown / NaN —-#: at most two per run), so a mean rise above one per run is unreachable, and the shader_v2 rubric+#: at most two per run), so a mean rise above one per run is unreachable, and the shader rubric#: caps those runs' score already. Articulated adds the pose sweep (capped per pair by the gate,#: ``MAX_PAIR_FINDINGS``) and a joint that moves the wrong way — the judge caps both#: (``joint_collision``), the veto must see them too.diff --git a/eval/bench/_loop_prompts.py b/eval/bench/_loop_prompts.pyindex 09b59f77..d1444a50 100644--- a/eval/bench/_loop_prompts.py+++ b/eval/bench/_loop_prompts.py@@ -47,7 +47,7 @@ FEATURE_POOL: tuple[str, ...] = ("night", "water", "interior", "particles")#: scene by construction, and the diagnosis would then blame the variant for the battery. The#: floor IS the scene_frames gate's threshold (``frame_metrics.DARK_MEAN_LUM``, imported, never a#: copy that drifts when the gate moves): an authored frame whose mean luminance is below it is a-#: ``dark_frame`` ERROR, capped by the scene_v2 rubric. The gate has no ceiling on the mean+#: ``dark_frame`` ERROR, capped by the scene rubric. The gate has no ceiling on the mean#: (``blown_frame`` is a fraction of pure-white pixels, ``frame_metrics.BLOWN_FRAC``), so the#: ceiling is the loop's own margin under the white point.LUMINANCE_FLOOR = DARK_MEAN_LUMdiff --git a/eval/bench/compare_backends.py b/eval/bench/compare_backends.pyindex fce44a19..5bbc9abc 100644--- a/eval/bench/compare_backends.py+++ b/eval/bench/compare_backends.py@@ -18,7 +18,7 @@ ArmsEvery arm ends with a ``src/model.py`` that is copied into a fresh eval workspaceand scored by the SAME fixed evaluator: BlenderRuntime lint+build → measure →-connectivity gate → 14-view ``render_glb`` → ``VlmJudge(static_object_v1, judge,+connectivity gate → 14-view ``render_glb`` → ``VlmJudge(static_object, judge,n_samples=2)`` whose acceptance checklist is the battery's ``must_have`` list.A failed build (or unparseable answer) scores 0 with the error recorded. Then apairwise arena (``PairwiseJudge``, same judge model, both orders) runs everydiff --git a/eval/bench/judge_calib_graphics.py b/eval/bench/judge_calib_graphics.pyindex 4f5a476c..c06fb279 100644--- a/eval/bench/judge_calib_graphics.py+++ b/eval/bench/judge_calib_graphics.py@@ -4,7 +4,7 @@ The loop the object track had (rubric → re-judge recorded rounds → compare aperson sees → revise) never ran for graphics. This script is that loop's measuring half:python bench/judge_calib_graphics.py --corpus <corpus.json> --eye <eye.json> \- --rubrics shader_v2 --model gemini:gemini-3.1-pro-preview --n 2 \+ --rubrics shader --model gemini:gemini-3.1-pro-preview --n 2 \--out bench/out/judge_calib_graphics* ``corpus.json``: ``[[score, run_dir, sheet_path, generator], ...]`` — every graphics run with a@@ -66,7 +66,7 @@ def main(argv: list[str] | None = None) -> int:ap = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])ap.add_argument("--corpus", required=True)ap.add_argument("--eye", required=True)- ap.add_argument("--rubrics", default="shader_v2")+ ap.add_argument("--rubrics", default="shader")ap.add_argument("--model", default="gemini:gemini-3.1-pro-preview")ap.add_argument("--n", type=int, default=2)ap.add_argument("--workers", type=int, default=4)diff --git a/eval/bench/prompts/refs_v2_graphics.yaml b/eval/bench/prompts/refs_v2_graphics.yamlindex 33517a19..6e3d0853 100644--- a/eval/bench/prompts/refs_v2_graphics.yaml+++ b/eval/bench/prompts/refs_v2_graphics.yaml@@ -2,7 +2,7 @@ name: refs_v2_graphicstrack: graphicslanguage: glsl_shaderdescription: 'Second turn of the graphics loop: the same aurora brief, with/without the two photos, 3 seeds per- arm, judged in-loop by shader_v2 with the light-phenomena cookbook (aeb49c0) and brief-selected chapters (137541e).'+ arm, judged in-loop by shader with the light-phenomena cookbook (aeb49c0) and brief-selected chapters (137541e).'prompts:- id: aurora_ref_1references:diff --git a/eval/bench/rejudge_offline.py b/eval/bench/rejudge_offline.pyindex e1a9a447..e427d7d1 100644--- a/eval/bench/rejudge_offline.py+++ b/eval/bench/rejudge_offline.py@@ -28,8 +28,8 @@ Cost: seconds of CPU. It never writes into a run.CLI::- python bench/rejudge_offline.py bench/out --rubric static_object_v1 --out scratch/replay- python bench/rejudge_offline.py bench/out --rubric static_object_v1 --identity # exit 1 on any drift, 2 when nothing was held+ python bench/rejudge_offline.py bench/out --rubric static_object --out scratch/replay+ python bench/rejudge_offline.py bench/out --rubric static_object --identity # exit 1 on any drift, 2 when nothing was held"""from __future__ import annotations@@ -261,7 +261,7 @@ def replay_corpus(roots: list[Path], *, rubric_name: str, rubric: Rubric | Nonedef main(argv: list[str] | None = None) -> int:ap = argparse.ArgumentParser(description=__doc__.split("\n\n")[0])ap.add_argument("roots", nargs="+", type=Path)- ap.add_argument("--rubric", default="static_object_v1", help="rubric NAME the stored verdicts must carry")+ ap.add_argument("--rubric", default="static_object", help="rubric NAME the stored verdicts must carry")ap.add_argument("--rubric-file", type=Path, default=None,help="score with THIS yaml instead of the installed rubric of the same name (a variant under test)")ap.add_argument("--kinds", default="", help="comma list of round kinds to keep (default all, incl. candidate)")diff --git a/eval/docs/COMPLEXITY.md b/eval/docs/COMPLEXITY.mdindex 7637ecf8..32071190 100644--- a/eval/docs/COMPLEXITY.md+++ b/eval/docs/COMPLEXITY.md@@ -207,8 +207,8 @@ generation-side levers belong to the other two owners of the complexity wave.`complexity` block on every record (with `plan_parts`, `parts_per_plan_part`and the per-round trail), a `cx 0.65` chip on every gallery card, acomplexity panel and a complexity sort in the gallery.-2. **`geometry_detail` was made countable** (F2) in `static_object_v1`,- `articulated_v1`, `asset_v1` and `reference_v1`:+2. **`geometry_detail` was made countable** (F2) in `static_object`,+ `articulated`, `asset` and `reference`:* a named list of **eight refinement kinds** (bevel/fillet, taper, curve,moulding, cut-out, wall thickness, surface relief, modelled hardware) in therubric notes, so "detail" is a count and not a mood;diff --git a/eval/docs/EVAL.md b/eval/docs/EVAL.mdindex 2353d447..0147a745 100644--- a/eval/docs/EVAL.md+++ b/eval/docs/EVAL.md@@ -137,7 +137,7 @@ Arms:Every arm ends with a `src/model.py` copied into a fresh eval workspace and scored bythe same `FixedEvaluator` (the cell's runtime lint+build → measure → connectivity →-14-view render + clay views → `VlmJudge(static_object_v1, --judge, n_samples=2)`). Its payload+14-view render + clay views → `VlmJudge(static_object, --judge, n_samples=2)`). Its payloadis the harness's `round_input` (`_fixed_eval.fixed_input`), so it is the in-run judge's for thesame round — clay montage, the GLB for the D48 slice channel, connectivity in the author's frame —except the checklist, the brief's `must_have`. (N83, 2026-09-24: before that it was hand-built@@ -312,7 +312,7 @@ Notes: the earlier criteria-first schema compressed flash to 0.6–0.7 (std 0.01on 2026-09-22), and the 0.072 round-to-round spread in the corpus is generation variance,not the judge. Untested: temperature 0.0 (brilliana measured 0.013vs 0.035 between 0.0 and 0.2 on 512 calls) — one more $8 battery.-* **scene_v2 (2026-10-14): the scene judge gets stricter and detail-aware; scene_v1 was frozen+* **scene (2026-10-14): the scene judge gets stricter and detail-aware; scene_v1 was frozenfor replays at that time and is no longer distributed.** Part of the argon scene-tuning program (D105): the owner asked for stricterconnectivity / contact standards, and the scene judge was the lenient half of them — theplacement gate measured floating / sunken / interpenetration while the rubric had one −0.06@@ -336,7 +336,7 @@ Notes: the earlier criteria-first schema compressed flash to 0.6–0.7 (std 0.01placement rows and the plan checks share one report, so a `no_fog` ERROR let every floatingclaim stand although the rows had measured none. Now a watched report is clean for a rulewhen it ran and has no ERROR of the rule's kinds and no WARN of those kinds at or over the- rule's `veto_warn_keys` line (scene_v2: `gap_m` 0.02 / `sunk_m` 0.10 / `overlap` 0.20 / any+ rule's `veto_warn_keys` line (scene: `gap_m` 0.02 / `sunk_m` 0.10 / `overlap` 0.20 / anyscale factor — a WARN-band hover at 0.08 m keeps the judge's tick, and so does a 3 cm one:the gap line is the INDOOR tolerance = `contact_m`, so a floating / unsupported WARN the gateraised is never vetoed; it was the outdoor 0.05 until 2026-10-15 and vetoed 2-5 cm indoor@@ -371,7 +371,7 @@ Notes: the earlier criteria-first schema compressed flash to 0.6–0.7 (std 0.01* **The A/B evaluator** (`eval/bench/_fixed_eval.py`) runs `placement_gate` on a scene cellbefore the judge, so the deterministic caps fire in a compare cell as in the loop.* Not measured yet: the loop's judge-calibration arm (`bench.loop judge-calib`) is the- within-sd / between-sd / SNR readout for scene_v2 under argon@medium; until it runs, the+ within-sd / between-sd / SNR readout for scene under argon@medium; until it runs, thenumbers above are the rubric's, not a battery's.* Never tune rubric text against the battery you report on; bump the rubric version(`*_v2`) instead and re-run.@@ -385,7 +385,7 @@ same contact sheets ("would a curator screenshot it / does it look like the thinThree aurora versions nobody would take for an aurora scored 0.78 / 0.94 / 0.94 with empty issuelists; opaque pastel discs for bokeh 0.92; a lifted purple wash for a nebula 0.82; a crispukiyo-e wave 0.59 under planner must items. The rubric scored the nouns of the brief being-present. The graphics loop fixes this (rubric `shader_v2`: likeness,+present. The graphics loop fixes this (rubric `shader`: likeness,tonal range, an artefact checklist; reference photos via `bench/refs/<id>/`; `LikenessJudge`)and the ledger of turns; `bench/judge_calib_graphics.py` re-judges the corpus under two rubricsagainst the eye file and is the gate for switching the track default. The rule from §6 holds:@@ -685,7 +685,7 @@ verbatim, `must_have` empty), our runs on the same prompts (generator**`api-agent:gemini:gemini-3.7-flash`** — the in-process arm, retired two days later by `66175ec`;`candidates=1`, `texture=false`; ≤ 3 rounds, on a 503-storm day), renders BOTH GLBs with ourrenderer, gates both, and judges both with one fixed pro judge-(`static_object_v1`, n=2) — twice: **visual only** (no gate text in the judge's context) and+(`static_object`, n=2) — twice: **visual only** (no gate text in the judge's context) and**gated** (the gate findings in the context, what the harness itself would say). The galleryGLBs are merged, un-welded exports whose "floating" findings are export artefacts as often asdefects, so the visual number is the fair headline. `bench/h2h_scene.py` (scenes) does the@@ -1195,7 +1195,7 @@ the original `off` roll missed, so the "drift" is the baseline's own vote varian`untextured_flat` votes overall: 8 → 5, the drop landing exactly on the adjudicated-false cases.New text: default grey / magenta placeholder (the untinted "nothing was assigned" look — a defecteven when lit), or UNLIT fills where differently-angled faces render the SAME brightness; "no-texture map" is never the test. `judge_prompt_hash` moves for `static_object_v1`, as any wording+texture map" is never the test. `judge_prompt_hash` moves for `static_object`, as any wordingchange does; scoring is untouched.### 2026-09-01 — sceneloop: the scene-track efficiency ledger, the first skills A/B (inconclusive), and two refuted hypothesesdiff --git a/eval/docs/PAPER_WRITING.md b/eval/docs/PAPER_WRITING.mdindex fff40486..a33e194e 100644--- a/eval/docs/PAPER_WRITING.md+++ b/eval/docs/PAPER_WRITING.md@@ -98,7 +98,7 @@ are marked as confounded wherever they appear below.Everything below is judged by a **fixed evaluator**: the delivered GLB re-rendered on a14-view rig and scored by `gemini:gemini-3.1-pro-preview` with `n_samples=3` against the-`articulated_v1` rubric. Arms are paired per prompt. `--pin-plan` runs the planner once+`articulated` rubric. Arms are paired per prompt. `--pin-plan` runs the planner onceand seeds both arms, removing the planner's spread from the difference.**A/A calibration (2026-09-02, `bench/out/aa_articulated`).** Two *identical* arms ondiff --git a/eval/llm/README.md b/eval/llm/README.mdindex 42f481d9..079ad00c 100644--- a/eval/llm/README.md+++ b/eval/llm/README.md@@ -123,7 +123,7 @@ three.js 40/40 — identical to the report's reference counts.## Reusing the harness judgeThe harness batteries (`../bench/prompts/*.yaml`, the ten in `config.HARNESS_BATTERIES`) have no ground truth; their quality score is the calibrated-Gemini VLM judge in `harness/codeverse3d/judges` (14-view rig, rubric `static_object_v1`, σ ≈ 0.03 at n=3).+Gemini VLM judge in `harness/codeverse3d/judges` (14-view rig, rubric `static_object`, σ ≈ 0.03 at n=3).`docs/inventory_harness_bench.md` §7 shows the minimal call sequence (`FixedEvaluator`) tojudge a GLB produced here without the agentic pipeline.diff --git a/eval/llm/docs/inventory_harness_bench.md b/eval/llm/docs/inventory_harness_bench.mdindex 8f3f10ec..b82a30d3 100644--- a/eval/llm/docs/inventory_harness_bench.md+++ b/eval/llm/docs/inventory_harness_bench.md@@ -31,7 +31,7 @@ Schema (`bench/run_bench.py:49-71`): `Battery{name, track, language, description- `codeverse3d/judges/vlm_judge.py` `VlmJudge(rubric, model_id, n_samples, temperature=0.2, thinking="low")`; default judge `gemini:gemini-3.1-pro-preview`. Variants `ReferenceJudge` (reference images + silhouette IoU), `LikenessJudge`, `PairwiseJudge` (`judges/pairwise.py`: both orderings, winner only when they agree).- Prompt (`judges/prompt_builder.py`): blind judge, observe-then-score, defect checklist; BRIEF + CONSTRAINTS + MEASUREMENTS + GATE FINDINGS; ≤5 labelled 2×2 montages (512 px tiles) + ≤2 detail crops; cross-section slices only on connectivity ERROR (D48).- Score (`judges/rubrics.py`, SCORING_VERSION=2): weighted criteria → majority-voted defects with penalties/caps → gate caps → `passed = overall ≥ threshold ∧ no floor ∧ no unverified must item`.-- Rubrics (`judges/rubrics/*.yaml`): `static_object_v1` (threshold 0.72; intent_fidelity .22, structure .18, geometry_detail .16, proportions .12, assembly_fit .12, materials .10, craftsmanship .10; 11 defects), `articulated_v1` (0.70), `scene_v1` (0.70), `shader_v2` (0.70; shader_v1 had Spearman 0.16 vs human), `asset_v1`, `reference_v1` (measured silhouette IoU criterion, weight .25).+- Rubrics (`judges/rubrics/*.yaml`): `static_object` (threshold 0.72; intent_fidelity .22, structure .18, geometry_detail .16, proportions .12, assembly_fit .12, materials .10, craftsmanship .10; 11 defects), `articulated` (0.70), `scene_v1` (0.70), `shader` (0.70; shader_v1 had Spearman 0.16 vs human), `asset`, `reference` (measured silhouette IoU criterion, weight .25).- View rig (`codeverse3d/conventions.py:87-102`): 14 views (high ring ×4 @ +30°, eye-level ×4, low ring ×4 @ −30°, top, bottom), 768² studio background; clay montage ×4. Graphics: frames at t = 0, 1, 2.5, 4, 6 s.- Calibration (`docs/EVAL.md` §6): pro σ(overall, n=3) **0.030**, flash σ **0.083**; fixed-order re-judge σ 0.035; paired A/A on identical arms: paired sd 0.16–0.20 → generation variance dominates; ±0.02 needs ~500 pairs. ~$0.15–0.20 per pro verdict.- Deterministic gates: per-language lint; build (`ok` requires non-empty GLB); `measure_glb` (bbox, tri_count, n_meshes, n_islands, per-part volume/watertight, ground_gap, materials, complexity vector); `check_connectivity` (floating part = ERROR, penetration WARN > 2 mm, ERROR > 10 mm); graphics `gl_frames` (NaN, black/blown, static, flicker, low_detail); `reference_silhouette` IoU; `must_have` → VLM acceptance items.diff --git a/eval/llm/docs/metrics.md b/eval/llm/docs/metrics.mdindex 356783c6..a5997942 100644--- a/eval/llm/docs/metrics.md+++ b/eval/llm/docs/metrics.md@@ -44,7 +44,7 @@ Sampling at T = 0.7, top-p 0.95, seeds 0..N−1 (`run_eval --samples N`).## 4. Rubric-judged suites (harness batteries, no GT)Execution rate is always reported. Quality needs a judge: the harness's calibrated VLM judge-(`harness/codeverse3d/judges`, Gemini pro, 14-view rig, `static_object_v1` rubric, threshold 0.72, σ ≈ 0.03 at+(`harness/codeverse3d/judges`, Gemini pro, 14-view rig, `static_object` rubric, threshold 0.72, σ ≈ 0.03 atn=3) is the reference implementation — see `docs/inventory_harness_bench.md` §2 for how to call it on aGLB. No judge is bundled here yet; if one is added it must be pairwise with swapped order and report κ against ahuman-labelled subset (LLM judges have position/length bias — survey §2e).diff --git a/eval/tests/conftest.py b/eval/tests/conftest.pyindex 3585dc28..9aaa3e75 100644--- a/eval/tests/conftest.py+++ b/eval/tests/conftest.py@@ -108,7 +108,7 @@ class FakeEvaluator:gate = GateReport(gate="connectivity", passed=score >= 0.4, findings=[] if score >= 0.4 else [GateFinding(gate="connectivity", severity=Severity.ERROR, target="Body", message="floating part")])out.gates.append(gate)- out.judgment = Judgment(rubric="static_object_v1", judge_backend="fake", scores={"intent_fidelity": score},+ out.judgment = Judgment(rubric="static_object", judge_backend="fake", scores={"intent_fidelity": score},overall=score, passed=score >= 0.72, n_samples=2, usage=Usage(cost_usd=0.01))return out@@ -141,7 +141,7 @@ def fake_run_track(score: float = 0.8):(ws.root / MODEL_FILE).write_text(GOOD.format(score=score))(ws.src / "parts").mkdir(exist_ok=True)(ws.src / "parts" / "legs.py").write_text("LEGS = 3\n")- loop_verdict = Judgment(rubric="static_object_v1", scores={}, overall=round(score - 0.1, 4), passed=False)+ loop_verdict = Judgment(rubric="static_object", scores={}, overall=round(score - 0.1, 4), passed=False)rec = RunRecord(spec=spec, workspace=str(ws.root), status=RunStatus.MAX_ROUNDS,rounds=[RoundRecord(index=0, kind="baseline", judgment=loop_verdict,build=BuildResult(ok=True, language="blender"))],diff --git a/eval/tests/test_articulated_compare.py b/eval/tests/test_articulated_compare.pyindex b9fa2a2f..86472c2a 100644--- a/eval/tests/test_articulated_compare.py+++ b/eval/tests/test_articulated_compare.py@@ -1,4 +1,4 @@-"""compare_backends on an articulated battery: two-file one-shot answers, articulated_v1 fixed judge."""+"""compare_backends on an articulated battery: two-file one-shot answers, articulated fixed judge."""from __future__ import annotations@@ -93,20 +93,20 @@ def test_rubric_for_is_the_one_track_to_rubric_mapping():for t in Track:assert rubric_for(t) == TRACK_INFO[t].rubricassert rubric_for(Spec(id="x", track=t, language=lang[t], prompt="x")) == TRACK_INFO[t].rubric- assert report.battery_rubric({"battery": {"track": "articulated_object"}}) == "articulated_v1"- assert report.battery_rubric({}) == "static_object_v1" == report.battery_rubric({"battery": {"track": "nope"}})+ assert report.battery_rubric({"battery": {"track": "articulated_object"}}) == "articulated"+ assert report.battery_rubric({}) == "static_object" == report.battery_rubric({"battery": {"track": "nope"}})# a static-battery evaluator still judges an articulated cell on the articulated rubric ...ev = FixedEvaluator("gemini:x", n_samples=1)- assert ev.judge_for(_urdf_spec()).rubric.name == "articulated_v1"+ assert ev.judge_for(_urdf_spec()).rubric.name == "articulated"assert ev.judge_for(_urdf_spec()) is ev.judge_for(_urdf_spec()), "one judge per rubric, cached"- assert ev.judge.rubric.name == "static_object_v1"+ assert ev.judge.rubric.name == "static_object"# ... unless the caller pinned a rubric for every cell- assert FixedEvaluator("gemini:x", rubric="asset_v1").judge_for(_urdf_spec()).rubric.name == "asset_v1"+ assert FixedEvaluator("gemini:x", rubric="asset").judge_for(_urdf_spec()).rubric.name == "asset"def test_articulated_cells_add_the_joint_sweep_and_the_pose_sheet(monkeypatch, tmp_path):"""evaluate() on an articulated spec: the sweep gate joins the gates, the pose views join the- renders, and the judge is the cell's (articulated_v1) — keyed on the SPEC's track."""+ renders, and the judge is the cell's (articulated) — keyed on the SPEC's track."""from types import SimpleNamespaceimport codeverse3d.spatial.connectivity as conn@@ -131,7 +131,7 @@ def test_articulated_cells_add_the_joint_sweep_and_the_pose_sheet(monkeypatch, tmonkeypatch.setattr(art, "default_joint_sweep", lambda ws, plan, out: (seen.__setitem__("plan", plan), (sweep, [pose]))[1])class _J:- rubric = SimpleNamespace(name="articulated_v1")+ rubric = SimpleNamespace(name="articulated")def judge(self, inp):seen["inp"] = inp@@ -161,6 +161,6 @@ def test_main_builds_the_evaluator_from_the_battery(monkeypatch, tmp_path):rc = cb.main(["--prompts", str(PROMPTS / "articulated_v2.yaml"), "--arms", "oneshot:gemini:x","--out", str(tmp_path / "out"), "--no-preflight", "--judge-samples", "3"])assert rc == 0- assert seen == {"track": Track.ARTICULATED_OBJECT, "language": Language.URDF_BLENDER, "rubric": "articulated_v1"}+ assert seen == {"track": Track.ARTICULATED_OBJECT, "language": Language.URDF_BLENDER, "rubric": "articulated"}b = Battery.load(PROMPTS / "articulated_v2.yaml")assert b.track is Track.ARTICULATED_OBJECT and all(p.must_have for p in b.prompts)diff --git a/eval/tests/test_compare_matrix.py b/eval/tests/test_compare_matrix.pyindex 3a4dca8d..6edc8ea1 100644--- a/eval/tests/test_compare_matrix.py+++ b/eval/tests/test_compare_matrix.py@@ -220,7 +220,7 @@ def test_every_cell_and_its_harness_run_open_a_ledger(tmp_path: Path):inner_eval = ev.evaluatedef evaluate(ws, spec): # the fixed evaluator buys a verdict- bill("judge:static_object_v1:r00:s0")+ bill("judge:static_object:r00:s0")return inner_eval(ws, spec)ev.evaluate = evaluate # type: ignore[method-assign]@@ -239,10 +239,10 @@ def test_every_cell_and_its_harness_run_open_a_ledger(tmp_path: Path):cells = {r.arm: Path(r.workspace) for r in rows}harness = cells["harness:agy:gemini-3.6-flash"]assert [r.label for r in load_ledger(harness / "run")] == ["baseline"]- assert [r.label for r in load_ledger(harness)] == ["judge:static_object_v1:r00:s0"]+ assert [r.label for r in load_ledger(harness)] == ["judge:static_object:r00:s0"]oneshot = cells["oneshot:claude-code"]assert not (oneshot / "run").exists()- assert [r.label for r in load_ledger(oneshot)] == ["judge:static_object_v1:r00:s0"]+ assert [r.label for r in load_ledger(oneshot)] == ["judge:static_object:r00:s0"]def test_no_resume_starts_the_harness_arm_fresh(tmp_path: Path):diff --git a/eval/tests/test_complexity_report.py b/eval/tests/test_complexity_report.pyindex 74b0c008..e2c8e6fc 100644--- a/eval/tests/test_complexity_report.py+++ b/eval/tests/test_complexity_report.py@@ -45,7 +45,7 @@ def _write_run(root: Path, slug: str, *, index: float, overall: float, detail: f"rounds": [{"index": 0, "kind": "baseline", "gates": [],"measurement": _measurement(index).model_dump(mode="json"),- "judgment": {"rubric": "static_object_v1", "overall": overall, "passed": overall >= 0.72,+ "judgment": {"rubric": "static_object", "overall": overall, "passed": overall >= 0.72,"scores": {c: detail for c in CR.CRITERIA}, "issues": []},}],}@@ -105,7 +105,7 @@ def test_row_recomputes_the_vector_when_the_record_has_none(tmp_path: Path) -> N"status": "max_rounds","total_usage": {"cost_usd": 1.0}, "rounds": [{"index": 0, "kind": "baseline", "gates": [],- "judgment": {"rubric": "static_object_v1", "overall": 0.5, "passed": False,+ "judgment": {"rubric": "static_object", "overall": 0.5, "passed": False,"scores": dict.fromkeys(CR.CRITERIA, 0.5), "issues": []}}]}))row = CR.row_for(run, "old_battery")assert row is not None and row["part_count"] == 1 and row["index"] > 0diff --git a/eval/tests/test_fixed_eval_language.py b/eval/tests/test_fixed_eval_language.pyindex 96bbad8f..5bd4a306 100644--- a/eval/tests/test_fixed_eval_language.py+++ b/eval/tests/test_fixed_eval_language.py@@ -108,7 +108,7 @@ def test_scene_cells_are_judged_on_their_cameras_and_orbit_frames(monkeypatch, tdef test_scene_cells_run_the_placement_gate_before_the_judge(monkeypatch, tmp_path):- """scene_v2 (2026-10-14): a scene cell's deterministic placement caps fire in an A/B exactly as in+ """scene (2026-10-14): a scene cell's deterministic placement caps fire in an A/B exactly as inthe loop — ``placement_gate`` over the build's census and the cell's plan.json, before the framesgate; a build whose census has no placement table (the fake above) adds no report."""import jsondiff --git a/eval/tests/test_infra_failures.py b/eval/tests/test_infra_failures.pyindex 0f00437d..4db51263 100644--- a/eval/tests/test_infra_failures.py+++ b/eval/tests/test_infra_failures.py@@ -64,8 +64,8 @@ def test_scene_pairwise_uses_the_same_track_rubric_as_fixed_evaluation(tmp_path,pairs = cb.run_pairwise(battery, cells, arms, tmp_path, cb.CompareOptions(),cb.CompareDeps(FakeEvaluator(), pairwise_judge=Judge))- assert seen == ["scene_v2"]- assert pairs[0].rubric == "scene_v2"+ assert seen == ["scene"]+ assert pairs[0].rubric == "scene"OUTAGES = ["ModelError: Gemini API error 503: This model is currently experiencing high demand.",diff --git a/eval/tests/test_replay.py b/eval/tests/test_replay.pyindex b38e8684..c8001a9e 100644--- a/eval/tests/test_replay.py+++ b/eval/tests/test_replay.py@@ -27,7 +27,7 @@ from codeverse3d.judges.rubrics import (load_rubric,)-R = load_rubric("static_object_v1")+R = load_rubric("static_object")ITEMS = [AcceptanceItem(id="M1", text="four legs", priority="must"),AcceptanceItem(id="M2", text="0.45 m tall", priority="must", how="measure"),AcceptanceItem(id="S1", text="bevelled edges", priority="should")]diff --git a/eval/tests/test_scene_stats.py b/eval/tests/test_scene_stats.pyindex 5d7a4a13..f32cc9c4 100644--- a/eval/tests/test_scene_stats.py+++ b/eval/tests/test_scene_stats.py@@ -52,7 +52,7 @@ def test_the_median_score_is_read_off_the_real_Judgment_shape(tmp_path: Path) ->def judged(name: str, overall: float) -> None:d = tmp_path / name / "run"d.mkdir(parents=True, exist_ok=True)- j = Judgment(rubric="scene_v2", scores={"a": overall}, overall=overall, passed=overall > 0.5)+ j = Judgment(rubric="scene", scores={"a": overall}, overall=overall, passed=overall > 0.5)(d / "record.json").write_text(json.dumps({"spec": {"track": "scene", "id": name},"plan": {"assets": []},diff --git a/harness/CLAUDE.md b/harness/CLAUDE.mdindex 12f47e2d..e9e2cf35 100644--- a/harness/CLAUDE.md+++ b/harness/CLAUDE.md@@ -77,7 +77,7 @@ push, with `set -o pipefail` so a `| tail` cannot swallow a red exit. The versi`gemini-3.7-flash` stays the cheap API model (single-shot / economy, flash judge — lenient; needsthe defect checklist for range). Judge default `gemini-3.1-pro-preview` (`Settings.default_judge`; tracks usen_samples=1); flash + `--n 3` is the cheap fallback, pro for calibration/eval. The argon loop judges with- `gemini:gemini-4-argon` at `judge.thinking=medium` (argon does not think at 'low'; D105) on rubric `scene_v2`.+ `gemini:gemini-4-argon` at `judge.thinking=medium` (argon does not think at 'low'; D105) on rubric `scene`.- Live runs under `runs/` (`e2e_*`): read-only reference material; never modify.- Every model call and vendor session is priced into the run's `telemetry/cost.jsonl` — the ONLY record ofmoney: `record.total_usage` is its sum at list price (D84). A run's minutes are its steps' time minusdiff --git a/harness/codeverse3d/addons/costreport/caching.py b/harness/codeverse3d/addons/costreport/caching.pyindex 5607470e..67739265 100644--- a/harness/codeverse3d/addons/costreport/caching.py+++ b/harness/codeverse3d/addons/costreport/caching.py@@ -52,7 +52,7 @@ class SessionCache:def session_key(label: str) -> str:"""Collapse a per-call label to its session: ``api-agent:refine_legs:t7`` →- ``refine_legs``, ``judge:static_object_v1:r02:s1`` → ``judge:static_object_v1``.+ ``refine_legs``, ``judge:static_object:r02:s1`` → ``judge:static_object``.A session is what shares a prompt prefix: every turn of one agent session,every sample of one verdict."""diff --git a/harness/codeverse3d/contracts/common.py b/harness/codeverse3d/contracts/common.pyindex f05706db..1b61de74 100644--- a/harness/codeverse3d/contracts/common.py+++ b/harness/codeverse3d/contracts/common.py@@ -51,7 +51,7 @@ class TrackInfo(BaseModel):model_config = ConfigDict(frozen=True)- rubric: str = Field(description="rubric name under judges/rubrics/, e.g. 'static_object_v1'")+ rubric: str = Field(description="rubric name under judges/rubrics/, e.g. 'static_object'")label: str = Field(description="human label for galleries / captions / reports")best_of_n: bool = Field(default=True, description="the baseline is one session that --candidates N can run N ""times side by side (a scene writes its baseline in the pre-round stages: no)")@@ -61,10 +61,10 @@ class TrackInfo(BaseModel):#: one registry row per trackTRACK_INFO: dict[Track, TrackInfo] = {- Track.STATIC_OBJECT: TrackInfo(rubric="static_object_v1", label="3D Objects"),- Track.ARTICULATED_OBJECT: TrackInfo(rubric="articulated_v1", label="Articulated Objects"),- Track.SCENE: TrackInfo(rubric="scene_v2", label="3D Scenes", best_of_n=False, likeness_refs=True),- Track.GRAPHICS: TrackInfo(rubric="shader_v2", label="Procedural Graphics", likeness_refs=True),+ Track.STATIC_OBJECT: TrackInfo(rubric="static_object", label="3D Objects"),+ Track.ARTICULATED_OBJECT: TrackInfo(rubric="articulated", label="Articulated Objects"),+ Track.SCENE: TrackInfo(rubric="scene", label="3D Scenes", best_of_n=False, likeness_refs=True),+ Track.GRAPHICS: TrackInfo(rubric="shader", label="Procedural Graphics", likeness_refs=True),}#: entry file inside the workspace per language — the file a build starts from.diff --git a/harness/codeverse3d/conventions.py b/harness/codeverse3d/conventions.pyindex 7ec0cc3e..2a1482fa 100644--- a/harness/codeverse3d/conventions.py+++ b/harness/codeverse3d/conventions.py@@ -190,7 +190,7 @@ def views_by_preference(views: Sequence[RenderView], names: Sequence[str]) -> li#: Scene overview rig (scenes sit on ground: aerial + eye level, no underside). The#: ``overview_*`` views reach the judge (the first ``render_scene.MAX_JUDGE_OVERVIEWS``);-#: ``overview_low_front`` (2026-10-14, scene_v2) is the CONTACT camera — elevation under 10°+#: ``overview_low_front`` (2026-10-14, scene) is the CONTACT camera — elevation under 10°#: puts it at ground level just outside the content (``orbit.mjs fitContactCamera``), the#: angle at which daylight under a base and a part through a neighbour are visible, and the#: tile the judge's contact crops are cut from. The ``eye_*`` views are diagnostics only.diff --git a/harness/codeverse3d/cost/context.py b/harness/codeverse3d/cost/context.pyindex b21e19d0..fe12964b 100644--- a/harness/codeverse3d/cost/context.py+++ b/harness/codeverse3d/cost/context.py@@ -183,7 +183,7 @@ def _stated(stage: Stage | None) -> Stage | None:def _round_from_label(label: str) -> int | None:- """``judge:static_object_v1:r02:s1`` → 2 (``None`` when the label has no round tag)."""+ """``judge:static_object:r02:s1`` → 2 (``None`` when the label has no round tag)."""for token in label.replace("/", ":").split(":"):t = token.strip()if len(t) >= 2 and t[0] == "r" and t[1:].isdigit():diff --git a/harness/codeverse3d/judges/base.py b/harness/codeverse3d/judges/base.pyindex dce99673..30f2e36b 100644--- a/harness/codeverse3d/judges/base.py+++ b/harness/codeverse3d/judges/base.py@@ -26,7 +26,7 @@ from codeverse3d.workspace import Workspace#: scene and graphics have none and are unaffected.SLICE_TRACKS = ("static_object", "articulated_object")#: the rubric of a static object judged against reference images: it adds the measured silhouette criterion-REFERENCE_RUBRIC = "reference_v1"+REFERENCE_RUBRIC = "reference"def run_rubric(spec: Spec, default: str = "") -> str:diff --git a/harness/codeverse3d/judges/pairwise.py b/harness/codeverse3d/judges/pairwise.pyindex 274fe9ba..380bdbbc 100644--- a/harness/codeverse3d/judges/pairwise.py+++ b/harness/codeverse3d/judges/pairwise.py@@ -91,7 +91,7 @@ class PairwiseJudge:# ------------------------------------------------------------------ APIdef compare(self, spec: Spec, renders_a: RenderSet, renders_b: RenderSet, *,- rubric: str | Rubric = "static_object_v1"+ rubric: str | Rubric = "static_object") -> PairwiseResult:"""The two orderings' requests are tagged ``pairwise:fwd`` / ``pairwise:swap``so logs — and deterministic fakes — can tell concurrent orderings apart."""diff --git a/harness/codeverse3d/judges/prompt_builder.py b/harness/codeverse3d/judges/prompt_builder.pyindex aa76e420..075c17ac 100644--- a/harness/codeverse3d/judges/prompt_builder.py+++ b/harness/codeverse3d/judges/prompt_builder.py@@ -281,7 +281,7 @@ def plan_detail_crops(shaded: list[RenderView], *, scene: bool, gates: Sequence["""The crops a verdict sends, ``limit`` = ``Settings.judge.detail_crops`` (0 = none).Objects: the centre crop of the best view at 2× (detail legibility) + the ground band of- the lowest view, cut to ``limit``. Scenes (scene_v2, 2026-10-14): the centre crop plus up+ the lowest view, cut to ``limit``. Scenes (scene, 2026-10-14): the centre crop plus upto ``limit`` CONTACT CROPS — close-ups around the placement findings the gate measured(``plan_contact_crops``), so the judge sees the daylight under a base at crop scaleinstead of reading the gate's sentence about it.@@ -456,7 +456,7 @@ def plan_contact_crops(shaded: list[RenderView], gates: Sequence[GateReport]) ->size weighted by the tile's elevation (``CONTACT_TILE_WEIGHT``: the ground-level contacttile first, the top-down one last); a tile where the box fills the frame needs no crop. Nobox, no camera, no tile that sees it — no crop: the verdict never dies on a crop. Added- 2026-10-14 (scene_v2) because the judge read the gate's floating sentences instead of the+ 2026-10-14 (scene) because the judge read the gate's floating sentences instead of thepictures: the crop puts the measured contact in front of it at the scale a person wouldcheck."""overviews = [v for v in shaded if v.name in SCENE_OVERVIEW_NAMES]@@ -754,7 +754,7 @@ def gates_section(gates: list[GateReport], *, max_errors: int = 12, max_warns: iblock (contacts, planned joins contact/open, floor gaps) the judge can cite instead.Without a ledger (older rounds) the text is the pre-ledger text byte for byte, which iswhat keeps ``3dcode judge <old-slug>`` comparable with the verdict it stored.- ``scene`` (scene_v2, 2026-10-14) is the scene track's own branch — ``_scene_gates_section``;+ ``scene`` (scene, 2026-10-14) is the scene track's own branch — ``_scene_gates_section``;the object text above is untouched by it."""if not gates:return "GATE FINDINGS: (no gates run)"@@ -811,7 +811,7 @@ PLACEMENT_WORST: dict[str, tuple[str, str]] = {def _scene_gates_section(gates: list[GateReport], *, max_errors: int, max_warns: int) -> str:- """The scene judge's gate facts (scene_v2): reports in ``SCENE_GATE_ORDER``, the ERRORs as+ """The scene judge's gate facts (scene): reports in ``SCENE_GATE_ORDER``, the ERRORs asrecorded, one MEASURED PLACEMENT line built from the scene_placement report's own counts(``placement_block``), and the WARNs grouped by (gate, kind) as the object ledger branchgroups them — eight flat lines used to cut a round's WARNs mid-list with every frame@@ -1125,7 +1125,7 @@ RIG_RULES: dict[str, str] = {# BRIEF's place, not one recipe (foreground prop + blended ground + fog), and nothing_moves is the# MEASURED MOTION line's — the rubric forbids judging motion by comparing tiles"scene_craft": "On the authored cameras also read the CRAFT of the picture, not only its contents: depth readable for what each shot frames, surfaces that read as the brief's materials rather than one flat colour, variety among repeated elements, the secondary detail the brief's place implies, and distance softened where the brief's weather would soften it — each judged against what THIS place would have, not a fixed recipe. The same camera at two times appears as separate tiles; nothing_moves is answered from the MEASURED MOTION line only, never by comparing tiles.",- # scene_v2 (2026-10-14): the high overview tiles lift room ceilings (scene_host.skyLayersBelow) and+ # scene (2026-10-14): the high overview tiles lift room ceilings (scene_host.skyLayersBelow) and# drop the fog; the low one is the contact camera; CONTACT CROPs are diagnostic close-ups"scene_contact": "The high overview_* tiles (overview_front_right / overview_back_left / overview_top) are rendered with the fog removed and room ceilings lifted: a missing roof, lid or haze in them is never a defect. overview_low_front stands at ground level just outside the content, where daylight under a base shows. CONTACT CROP images are diagnostic close-ups cut from an overview tile around a placement finding the gate measured (the label names the asset, the kind and the number): judge the contact they show — base meets ground, part through neighbour — never their composition, lighting or dressing. Every floating / sunken / clipping / scale claim must name its view and its asset.","object": "All views show the same object. Use top + low views for symmetry, footprint and ground contact.",diff --git a/harness/codeverse3d/judges/rubrics.py b/harness/codeverse3d/judges/rubrics.pyindex 66d2633a..dcaf9cf1 100644--- a/harness/codeverse3d/judges/rubrics.py+++ b/harness/codeverse3d/judges/rubrics.py@@ -153,7 +153,7 @@ class CapRule(BaseModel):default_factory=dict,description="gate rules only: finding.data key → magnitude at or above which a WARN of the rule's kinds ""keeps the judge's claim alive (measured_absent); empty = the penetration default, "- "{depth_m: VETO_PENETRATION_DEPTH_M}. scene_v2 states {gap_m: 0.02} / {sunk_m: 0.10} / "+ "{depth_m: VETO_PENETRATION_DEPTH_M}. scene states {gap_m: 0.02} / {sunk_m: 0.10} / ""{overlap: 0.20}; 0.0 = any WARN of the kind that carries the key",)note: str = ""@@ -584,7 +584,7 @@ def _rule_gates_clean(rule: CapRule, gates: list[GateReport]) -> bool:Scoped to the rule's kinds since scoring version 4 (2026-10-14): a failed report used tocancel the veto whatever it failed on, and on the scene track the placement rows and theplan checks share ONE report — a ``no_fog`` or ``missing_content`` ERROR let every- "floating asset" claim stand although the rows had measured none (scene_v2 splits the+ "floating asset" claim stand although the rows had measured none (scene splits theplacement caps per kind for the same reason). A failed report that names no ERROR findingat all cannot say what failed and still counts as not clean."""watched = [g for g in gates if fnmatch(g.gate, rule.gate) or any(fnmatch(f.gate, rule.gate) for f in g.findings)]@@ -733,7 +733,7 @@ drift to report, not a fault.3 — 2026-09-24: a gate cap rule's ``kinds`` match the finding's typed ``data.kind`` /``data.code``; the message only when it has neither, the target never (N50). A replayof the 1 229 recorded judged rounds moved no cap and no veto.-4 — 2026-10-14 (scene_v2): the measured-absent veto is scoped to the rule's ``kinds`` — a+4 — 2026-10-14 (scene): the measured-absent veto is scoped to the rule's ``kinds`` — awatched report that failed on an unrelated kind no longer cancels it — and the WARNmagnitude that keeps a claim alive is the rule's ``veto_warn_keys`` (the object rubricskeep the ``depth_m`` ≥ 8 mm default, so their verdicts replay unchanged unless thediff --git a/harness/codeverse3d/judges/rubrics/articulated_v1.yaml b/harness/codeverse3d/judges/rubrics/articulated.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/articulated_v1.yamlrename to harness/codeverse3d/judges/rubrics/articulated.yamlindex c0093725..79e42bd8 100644--- a/harness/codeverse3d/judges/rubrics/articulated_v1.yaml+++ b/harness/codeverse3d/judges/rubrics/articulated.yaml@@ -1,4 +1,4 @@-name: articulated_v1+name: articulatedversion: 1track_hint: articulated_objectpass_threshold: 0.7@@ -6,7 +6,7 @@ description: 'Rubric for an ARTICULATED object (cabinet with doors/drawers, laptjudged on the rest-pose render rig PLUS the POSE SHEET: a harness-made grid whose rows are poses (rest, then eachjoint at a swept value, labelled ''<joint>@<value>'') and whose columns are views (3/4, back 3/4, front), withoptional 2x2 montages of pose_* tiles. The pose sheet is REQUIRED: without it the overall is capped (see caps)- because articulation cannot be judged. Static criteria are inherited from static_object_v1 at reduced weight;+ because articulation cannot be judged. Static criteria are inherited from static_object at reduced weight;articulation criteria score whether the joints are real, placed right and move plausibly.'extra_instructions: 'POSE tiles (the POSE SHEET image and any tile whose label starts with ''pose_'' or names ajoint@value) show the SAME object with joints moved by the harness (forward kinematics from the authored URDF);@@ -212,7 +212,7 @@ caps:when: acceptancecap: 0.6graded: true- note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object_v1); pass/fail unchanged+ note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object); pass/fail unchanged- id: unattached_linkwhen: gategate: '*'diff --git a/harness/codeverse3d/judges/rubrics/asset_v1.yaml b/harness/codeverse3d/judges/rubrics/asset.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/asset_v1.yamlrename to harness/codeverse3d/judges/rubrics/asset.yamlindex bce9125a..1e71182a 100644--- a/harness/codeverse3d/judges/rubrics/asset_v1.yaml+++ b/harness/codeverse3d/judges/rubrics/asset.yaml@@ -1,4 +1,4 @@-name: asset_v1+name: assetversion: 1track_hint: assetpass_threshold: 0.70@@ -143,7 +143,7 @@ caps:when: acceptancecap: 0.6graded: true- note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object_v1); pass/fail unchanged+ note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object); pass/fail unchanged# Binary checklist (judge answers, code scores: penalty subtracted, cap applied).defects:diff --git a/harness/codeverse3d/judges/rubrics/reference_v1.yaml b/harness/codeverse3d/judges/rubrics/reference.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/reference_v1.yamlrename to harness/codeverse3d/judges/rubrics/reference.yamlindex e9c0b699..c3779324 100644--- a/harness/codeverse3d/judges/rubrics/reference_v1.yaml+++ b/harness/codeverse3d/judges/rubrics/reference.yaml@@ -1,4 +1,4 @@-name: reference_v1+name: referenceversion: 1track_hint: referencepass_threshold: 0.70@@ -125,7 +125,7 @@ caps:when: acceptancecap: 0.6graded: true- note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object_v1); pass/fail unchanged+ note: 0.6 + 0.4 * verified/total must items (2026-08-30, as static_object); pass/fail unchanged# Binary checklist (judge answers, code scores: penalty subtracted, cap applied).defects:diff --git a/harness/codeverse3d/judges/rubrics/scene_v2.yaml b/harness/codeverse3d/judges/rubrics/scene.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/scene_v2.yamlrename to harness/codeverse3d/judges/rubrics/scene.yamlindex d37c1feb..cc854f06 100644--- a/harness/codeverse3d/judges/rubrics/scene_v2.yaml+++ b/harness/codeverse3d/judges/rubrics/scene.yaml@@ -1,4 +1,4 @@-name: scene_v2+name: sceneversion: 2track_hint: scenepass_threshold: 0.70diff --git a/harness/codeverse3d/judges/rubrics/shader_v2.yaml b/harness/codeverse3d/judges/rubrics/shader.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/shader_v2.yamlrename to harness/codeverse3d/judges/rubrics/shader.yamlindex 186bf5a6..a22751a6 100644--- a/harness/codeverse3d/judges/rubrics/shader_v2.yaml+++ b/harness/codeverse3d/judges/rubrics/shader.yaml@@ -1,4 +1,4 @@-name: shader_v2+name: shaderversion: 2track_hint: graphicspass_threshold: 0.70diff --git a/harness/codeverse3d/judges/rubrics/static_object_v1.yaml b/harness/codeverse3d/judges/rubrics/static_object.yamlsimilarity index 99%rename from harness/codeverse3d/judges/rubrics/static_object_v1.yamlrename to harness/codeverse3d/judges/rubrics/static_object.yamlindex bcb0bd55..1cddc763 100644--- a/harness/codeverse3d/judges/rubrics/static_object_v1.yaml+++ b/harness/codeverse3d/judges/rubrics/static_object.yaml@@ -1,4 +1,4 @@-name: static_object_v1+name: static_objectversion: 1track_hint: static_objectpass_threshold: 0.72diff --git a/harness/codeverse3d/judges/vlm_judge.py b/harness/codeverse3d/judges/vlm_judge.pyindex 3a0cbe6b..f80e2dc9 100644--- a/harness/codeverse3d/judges/vlm_judge.py+++ b/harness/codeverse3d/judges/vlm_judge.py@@ -82,7 +82,7 @@ THINKING_LEVELS: tuple[str, ...] = ("off", "low", "medium", "high")#: Per-track montage ceilings under the ``Settings.judge.montages`` dial. D47's 3 → 5 was#: measured on the OBJECT rig only (docs/DECISIONS.md D47); graphics keeps the pre-D47 3 until a-#: battery measures it. The scene rig moved 3 → 4 with scene_v2 (2026-10-14): the fourth+#: battery measures it. The scene rig moved 3 → 4 with scene (2026-10-14): the fourth#: overview (``overview_low_front``, the contact camera) plus the t=1.5 tiles need a third#: grid on a three-camera plan and a fourth on a four-camera one — a cap of 3 silently#: dropped the low tile the contact crops are cut from. An explicit ``max_montages`` wins.@@ -108,7 +108,7 @@ class VlmJudge:def __init__(self,- rubric: str | Rubric = "static_object_v1",+ rubric: str | Rubric = "static_object",model_id: str | None = None,n_samples: int = 1,temperature: float = 0.2,@@ -390,7 +390,7 @@ class ReferenceJudge(VlmJudge):n_samples: int = 1,temperature: float = 0.2,*,- rubric: str | Rubric = "reference_v1",+ rubric: str | Rubric = "reference",silhouette_fn: CompareFn | None = None,diff: bool = True,diff_model: Any | None = None,@@ -533,7 +533,7 @@ class LikenessJudge(VlmJudge):name = "likeness"def __init__(self, model_id: str | None = None, n_samples: int = 1, temperature: float = 0.2, *,- rubric: str | Rubric = "shader_v2", max_refs: int = 3, **kwargs: Any):+ rubric: str | Rubric = "shader", max_refs: int = 3, **kwargs: Any):super().__init__(rubric, model_id, n_samples, temperature, **kwargs)self.max_refs = max(1, int(max_refs))diff --git a/harness/codeverse3d/prompts/glsl_shader/contract.md b/harness/codeverse3d/prompts/glsl_shader/contract.mdindex c6c8ab15..16408692 100644--- a/harness/codeverse3d/prompts/glsl_shader/contract.md+++ b/harness/codeverse3d/prompts/glsl_shader/contract.md@@ -39,7 +39,7 @@ it writes PNGs, so what you compute is what you see). `u_prev` / `u_buffer_a` a`fragCoord / u_resolution` (0..1). Circles must stay round at the plan's resolution — the judge's`stretched_aspect` defect (−0.06): e.g. normalise both axes by `u_resolution.y`.-## What is measured (the `gl_frames` gate on the judged frames, then the `shader_v2` judge)+## What is measured (the `gl_frames` gate on the judged frames, then the `shader` judge)* A NaN / Inf pixel in any frame → ERROR, score cap 0. NaN sources in GLSL: division by zero, sqrt / pow /log of a negative, normalize of a zero vector, acos / asin outside [-1, 1].* Every frame essentially black (more than 97 % of its pixels under luminance 0.03) or blown (more than 97 %diff --git a/harness/codeverse3d/prompts/glsl_shader/system.md b/harness/codeverse3d/prompts/glsl_shader/system.mdindex 62d187f8..2aa12897 100644--- a/harness/codeverse3d/prompts/glsl_shader/system.md+++ b/harness/codeverse3d/prompts/glsl_shader/system.md@@ -29,7 +29,7 @@ t = 0, 1, 2.5, 4, 6 s must differ when motion is planned (`static`, cap 0.5) and(`flicker`, cap 0.7); no NaN pixel (cap 0); no black or blown frame (cap 0.5); no lint ERROR(cap 0.6). `u_frame` is not a clock — the sampled times are not evenly spaced.-What the judge (`shader_v2`) names, and what each costs:+What the judge (`shader`) names, and what each costs:* **Likeness (0.26)** — the thing drawn with the wrong physics scores 0.4 however completeits inventory: evenly spaced bars standing in for curtains, rays, flames or wavesdiff --git a/harness/codeverse3d/skills/c3d-glsl-craft/SKILL.md b/harness/codeverse3d/skills/c3d-glsl-craft/SKILL.mdindex b45a24a1..d870bfeb 100644--- a/harness/codeverse3d/skills/c3d-glsl-craft/SKILL.md+++ b/harness/codeverse3d/skills/c3d-glsl-craft/SKILL.md@@ -1,11 +1,11 @@---name: c3d-glsl-craft-description: "Use when writing, refining or repairing a glsl_shader fragment shader, or when gl_frames reports static, flicker or low detail. What the five sampled frames measure and where the real operating band is, what the shader_v2 judge names and charges for, and a short list of strong optional practices behind its recorded complaints: structure that does not alias, motion from the mechanism instead of a uv scroll, light as a path through a volume, exposure before colour."+description: "Use when writing, refining or repairing a glsl_shader fragment shader, or when gl_frames reports static, flicker or low detail. What the five sampled frames measure and where the real operating band is, what the shader judge names and charges for, and a short list of strong optional practices behind its recorded complaints: structure that does not alias, motion from the mechanism instead of a uv scroll, light as a path through a volume, exposure before colour."license: Apache-2.0compatibility: GLSL 330 core via moderngl, headless. One image pass (src/shader.frag), an optional src/common.glsl and one feedback pass (src/buffer_a.frag). The harness declares the uniforms; the contract lists them.metadata:evidence: mixed- evidence_note: "Gate thresholds, caps and rubric weights are read from live code (spatial/frame_stats.py, judges/rubrics/shader_v2.yaml) and pinned in _claims. The corpus is thin: 9 graded glsl_shader runs mined 2026-08-25 under the previous rubric, so every count below names its n and no rate is generalised."+ evidence_note: "Gate thresholds, caps and rubric weights are read from live code (spatial/frame_stats.py, judges/rubrics/shader.yaml) and pinned in _claims. The corpus is thin: 9 graded glsl_shader runs mined 2026-08-25 under the previous rubric, so every count below names its n and no rate is generalised."verified: "2026-10-15"corpus: "9 graded glsl_shader runs (18 judged rounds) + 5 opengl_python runs, mined 2026-08-25"target_metric: "mean_edge_density"@@ -33,7 +33,7 @@ them is a recipe — the brief decides what is on screen and how it is built.* `nan` — any NaN / Inf pixel (ERROR, cap 0)Over the 9 graded glsl runs mined 2026-08-25 the gate fired nothing while the judge sat at-0.55-0.65: clearing it proves nothing. The judge is `shader_v2`: likeness 0.26,+0.55-0.65: clearing it proves nothing. The judge is `shader`: likeness 0.26,depth_and_detail 0.16, light_and_tone 0.16, motion_quality 0.14, composition 0.10,technical_cleanliness 0.10, craft 0.08, plus the defect list the system prompt carries.One hint, and only a hint at n = 9: the harness's colourfulness metric (`gl_frames` printsdiff --git a/harness/codeverse3d/skills/c3d-scene-composition/SKILL.md b/harness/codeverse3d/skills/c3d-scene-composition/SKILL.mdindex 864a5fc7..ca328671 100644--- a/harness/codeverse3d/skills/c3d-scene-composition/SKILL.md+++ b/harness/codeverse3d/skills/c3d-scene-composition/SKILL.md@@ -5,7 +5,7 @@ license: Apache-2.0compatibility: track scene, language scene_threejs (three r182, headless Chrome, the harness builds the cameras from your plain camera objects).metadata:evidence: mixed- evidence_note: "Camera checks, score caps and the backdrop classifier are read from live code (spatial/frame_metrics.py, judges/rubrics/scene_v2.yaml, runtime_js/lib/backdrop.mjs, runtime_js/lib/scene_host.mjs). The corpus behind them is thin: only 3 graded scene runs exist, so every count below names its n."+ evidence_note: "Camera checks, score caps and the backdrop classifier are read from live code (spatial/frame_metrics.py, judges/rubrics/scene.yaml, runtime_js/lib/backdrop.mjs, runtime_js/lib/scene_host.mjs). The corpus behind them is thin: only 3 graded scene runs exist, so every count below names its n."verified: "2026-10-15"corpus: "3 graded scene_threejs runs with gate artefacts, mined 2026-08-25; 60 argon iteration-0 scenes re-read 2026-10-15 for the sameness audit"target_metric: "camera_placement_findings"@@ -70,7 +70,7 @@ Three consequences to design around:## Cameras are measured, and the numbers are hard`codeverse3d/spatial/frame_metrics.py` runs on the authored cameras, where every finding is an-**ERROR** (the harness's own orbit rig only gets WARNs). `judges/rubrics/scene_v2.yaml`+**ERROR** (the harness's own orbit rig only gets WARNs). `judges/rubrics/scene.yaml`turns those ERRORs into score ceilings:| condition | constant | finding | cap on the run |diff --git a/harness/codeverse3d/skills/c3d-scene-composition/references/measured-cameras.md b/harness/codeverse3d/skills/c3d-scene-composition/references/measured-cameras.mdindex a4c9a6b8..be0c9c9d 100644--- a/harness/codeverse3d/skills/c3d-scene-composition/references/measured-cameras.md+++ b/harness/codeverse3d/skills/c3d-scene-composition/references/measured-cameras.md@@ -1,7 +1,7 @@# The camera instruments, and what each finding asks you to changeCompanion to `c3d-scene-composition`. Constants read from-`codeverse3d/spatial/frame_metrics.py`, `codeverse3d/judges/rubrics/scene_v2.yaml`,+`codeverse3d/spatial/frame_metrics.py`, `codeverse3d/judges/rubrics/scene.yaml`,`runtime_js/lib/backdrop.mjs`, `runtime_js/lib/host_census.mjs` and`runtime_js/lib/scene_host.mjs` on 2026-08-25, re-read 2026-10-15.diff --git a/harness/codeverse3d/spatial/frame_stats.py b/harness/codeverse3d/spatial/frame_stats.pyindex 824b93fc..3b5c266a 100644--- a/harness/codeverse3d/spatial/frame_stats.py+++ b/harness/codeverse3d/spatial/frame_stats.py@@ -133,7 +133,7 @@ def sequence_stats(frames: Sequence[tuple[float, str | Path]], *, nan_counts: Seseq.mean_diff = float(np.mean(diffs))seq.max_diff = float(np.max(diffs))seq.n_duplicates = sum(1 for d in diffs if d < DUPLICATE_DIFF)- # MAX, not mean: 'static' is a scoring cap (shader_v2.yaml static_frames), so it must+ # MAX, not mean: 'static' is a scoring cap (shader.yaml static_frames), so it must# require that NO sampled pair changed. With the mean, an effect whose motion is# concentrated between two of the five sampled times (0/1/2.5/4/6 s) is averaged down# by the three quiet pairs and convicted while animating.diff --git a/harness/codeverse3d/spatial/render_blender.py b/harness/codeverse3d/spatial/render_blender.pyindex 5e8b35ee..96d155d7 100644--- a/harness/codeverse3d/spatial/render_blender.py+++ b/harness/codeverse3d/spatial/render_blender.py@@ -188,7 +188,7 @@ def fit_orbit(census_path: Path, views: Sequence[ViewPreset], *, width: int, heidef _pixel_orbit(specs: list[dict[str, Any]]) -> set[str]:"""The rig views that get Blender pixels: the first ``MAX_JUDGE_OVERVIEWS`` overviews — thesame four tiles ``render_scene.select_judge_views`` hands the judge (the three high views- and the ground-level ``overview_low_front`` since scene_v2); the eye-level rig is measured+ and the ground-level ``overview_low_front`` since scene); the eye-level rig is measuredby the geometry pass and never rendered."""return set([str(o["name"]) for o in specs if str(o["name"]).startswith("overview")][:MAX_JUDGE_OVERVIEWS])diff --git a/harness/codeverse3d/spatial/render_scene.py b/harness/codeverse3d/spatial/render_scene.pyindex b0812681..39b026f0 100644--- a/harness/codeverse3d/spatial/render_scene.py+++ b/harness/codeverse3d/spatial/render_scene.py@@ -42,7 +42,7 @@ log = logging.getLogger(__name__)#: views the judge sees (contact sheet + individual images are built from these)JUDGE_MAX_VIEWS = 10#: how many overview-rig tiles the judge gets (layout / grounding X-ray; the eye-level-#: rig views are diagnostics only and never reach the judge). 4 since scene_v2+#: rig views are diagnostics only and never reach the judge). 4 since scene#: (2026-10-14): the three high views plus ``overview_low_front``, the ground-level contact#: camera the judge's contact crops are cut from (``conventions.SCENE_VIEWS``).MAX_JUDGE_OVERVIEWS = 4diff --git a/harness/codeverse3d/spatial/scene_placement.py b/harness/codeverse3d/spatial/scene_placement.pyindex 92dda1b0..02151a21 100644--- a/harness/codeverse3d/spatial/scene_placement.py+++ b/harness/codeverse3d/spatial/scene_placement.py@@ -60,7 +60,7 @@ GATE = "scene_placement"#: SUNK_M / OVERLAP_MIN_FRAC): those decide what is MEASURED, these decide severities.MAX_FINDINGS_PER_KIND = 8#: ``free_exempt`` needs this many ``placement='free'`` rows before their share is a finding: one-#: hanging lantern among two placed things is 50 % and would cap a correct scene (scene_v2 caps a+#: hanging lantern among two placed things is 50 % and would cap a correct scene (scene caps a#: free_exempt ERROR at 0.75). The case it exists for — w1_tearoom, argon, 2026-10-13 — tagged 17.FREE_EXEMPT_MIN_ROWS = 2#: ``cantilevered`` needs this many foot columns resting on at most this many: a crate seated by itsdiff --git a/harness/codeverse3d/texturing/generate.py b/harness/codeverse3d/texturing/generate.pyindex 1f298b00..4be0c2a3 100644--- a/harness/codeverse3d/texturing/generate.py+++ b/harness/codeverse3d/texturing/generate.py@@ -346,7 +346,7 @@ def judge_gate(return resres.overall_before, res.overall_after = float(jb.overall), float(ja.overall)res.delta = round(res.overall_after - res.overall_before, 4)- crit = load_rubric(ja.rubric).materials_criterion # "" (articulated_v1): Δoverall alone decides+ crit = load_rubric(ja.rubric).materials_criterion # "" (articulated): Δoverall alone decidesres.materials_criterion = critif crit:res.materials_before = float(jb.scores.get(crit, 0.0))diff --git a/harness/codeverse3d/texturing/run.py b/harness/codeverse3d/texturing/run.pyindex 3ad6f448..4a2e4211 100644--- a/harness/codeverse3d/texturing/run.py+++ b/harness/codeverse3d/texturing/run.py@@ -261,7 +261,7 @@ def _make_judge(spec: Spec, rubric: str | None, judge_model_id: str | None) -> Afrom codeverse3d.judges.vlm_judge import VlmJudge# the track's own rubric even for a --image run, on purpose: this plain VlmJudge has no reference diff, so- # reference_v1's measured silhouette_match could not be scored, and textures never move a silhouette anyway+ # reference's measured silhouette_match could not be scored, and textures never move a silhouette anywayreturn VlmJudge(rubric=rubric or TRACK_INFO[spec.track].rubric, model_id=judge_model_id or spec.backends.judge,n_samples=1, label="texture_gate")diff --git a/harness/codeverse3d/tracks/common.py b/harness/codeverse3d/tracks/common.pyindex 77049602..21f41c0f 100644--- a/harness/codeverse3d/tracks/common.py+++ b/harness/codeverse3d/tracks/common.py@@ -59,11 +59,11 @@ class Services:def judge(self, rubric: str, model_id: str, n_samples: int = 1) -> Any:return _import("codeverse3d.judges.vlm_judge", "VlmJudge")(rubric=rubric, model_id=model_id, n_samples=n_samples)- def reference_judge(self, model_id: str, n_samples: int = 1, rubric: str = "reference_v1") -> Any:+ def reference_judge(self, model_id: str, n_samples: int = 1, rubric: str = "reference") -> Any:"""Image-conditioned judge (spec has reference images): renders + references + silhouette IoU."""return _import("codeverse3d.judges.vlm_judge", "ReferenceJudge")(model_id=model_id, n_samples=n_samples, rubric=rubric)- def likeness_judge(self, model_id: str, n_samples: int = 1, rubric: str = "shader_v2") -> Any:+ def likeness_judge(self, model_id: str, n_samples: int = 1, rubric: str = "shader") -> Any:"""Reference photos beside the frames, no silhouette (graphics / scene with images)."""return _import("codeverse3d.judges.vlm_judge", "LikenessJudge")(model_id=model_id, n_samples=n_samples, rubric=rubric)diff --git a/harness/codeverse3d/tracks/graphics.py b/harness/codeverse3d/tracks/graphics.pyindex cb08c9de..719d4086 100644--- a/harness/codeverse3d/tracks/graphics.py+++ b/harness/codeverse3d/tracks/graphics.py@@ -4,7 +4,7 @@ Languages: glsl_shader (Shadertoy-style fragment shader) · opengl_python (rawmoderngl program). Reuses ``BaseTrack`` (lifecycle) and ``run_round`` (steps)unchanged; the track-specific pieces are the prompt context (no 3D frame) and therender step (the sampled frames + contact sheet as the RenderSet-the ``shader_v2`` judge sees); the ``gl_frames`` gate is the build's own+the ``shader`` judge sees); the ``gl_frames`` gate is the build's own(``BuildResult.gates``). Planning is ``tracks/planner.py``'s, like every track's(template, worked example, T = 0.5, acceptance: all dispatched on the track there).Refinement is always one whole-program task.diff --git a/harness/codeverse3d/tracks/scene.py b/harness/codeverse3d/tracks/scene.pyindex ba716d88..f1368663 100644--- a/harness/codeverse3d/tracks/scene.py+++ b/harness/codeverse3d/tracks/scene.py@@ -3,7 +3,7 @@Languages: every scene language, through its ``SceneRuntime`` (D101). Generation is staged (each stage cached by``StageRunner`` for resume); the round loop then builds (probe + shaders — thebuild's two gate reports join the round's gates),-renders authored cameras + orbit at t=0 and t=1.5, judges with ``scene_v2``+renders authored cameras + orbit at t=0 and t=1.5, judges with ``scene``(console/shader errors become gate ERRORS) and dispatches refine tasks perzone / asset / env / camera, in parallel when file-disjoint.diff --git a/harness/codeverse3d/tracks/scene_assets.py b/harness/codeverse3d/tracks/scene_assets.pyindex 1f4c95a1..75a4f820 100644--- a/harness/codeverse3d/tracks/scene_assets.py+++ b/harness/codeverse3d/tracks/scene_assets.py@@ -78,7 +78,7 @@ ASSET_AGENT_TIMEOUT_S = 420#: ran the full 420 s and held the stage to 10.9 min — 28 % of the run for one asset, and#: round 0 did not start until 18.9 min.ASSET_SESSION_SHARE = 0.15-ASSET_RUBRIC = "asset_v1"+ASSET_RUBRIC = "asset"#: the asset kinds that are a MODULE in the scene workspace (one factory per file, placed by the#: zones); ``blender_glb`` is the hero, built as an object and shipped as a GLBMODULE_KINDS: frozenset[str] = frozenset({"threejs", "bpy"})@@ -440,7 +440,7 @@ def build_blender_asset(ctx: RunContext, asset: AssetPlan, *, judge: bool) -> As09-05): 15 first sessions killed at the asset timeout (booked at $0 — a killed CLIreports no usage), 12 GLBs one joined mesh with vertex paint (the one-part plan plusthe Blender contract's "one object per plan part" leave no other way to colour it, so- wood and iron share one roughness), median asset_v1 0.57 with 2 of 18 passing, and+ wood and iron share one roughness), median asset 0.57 with 2 of 18 passing, andevery fix prompt reading "Style of the whole scene: (none)" because the sub-spec'sprompt IS the asset sheet."""@@ -670,7 +670,7 @@ def _judge_and_fix(ctx: RunContext, asset: AssetPlan, result: AssetResult, rendelanguage: Language, after_fix: Callable[[RunContext], bool] | None = None,measure_fn: Callable[[], Any] | None = None,snapshot: Callable[[], Callable[[], None]] | None = None) -> AssetResult:- """Quick-sheet judge with asset_v1 (n_samples=1); ONE fix pass when below threshold,+ """Quick-sheet judge with asset (n_samples=1); ONE fix pass when below threshold,then the fix is checked (``after_fix(gen_ctx)``) and JUDGED AGAIN — ``score`` describesthe asset that ships, ``score_before`` the one that was fixed. The fix climbs thesame cheap rung as the generation: one chat call when the run has a chat modeldiff --git a/harness/docs/ARCHITECTURE.md b/harness/docs/ARCHITECTURE.mdindex b56f242c..31b7d13f 100644--- a/harness/docs/ARCHITECTURE.md+++ b/harness/docs/ARCHITECTURE.md@@ -449,10 +449,10 @@ must-acceptance and`missing_views` rules apply on top; degraded verdicts are glitches, not scores.`PairwiseJudge` (position-swapped, tie on disagreement; only `3dcode pick --by pairwise` asks it); `ReferenceJudge` forimage-conditioned specs.-Rubrics (pass threshold 0.70 unless named): `static_object_v1` (0.72), `articulated_v1`-(requires pose views: the `missing_pose_sheet` cap), `scene_v2` (D105: placement caps floating 0.60 / sunken 0.65 /+Rubrics (pass threshold 0.70 unless named): `static_object` (0.72), `articulated`+(requires pose views: the `missing_pose_sheet` cap), `scene` (D105: placement caps floating 0.60 / sunken 0.65 /interpenetration 0.70 / scale 0.75 / free_exempt 0.75 measured by `scene_placement`, frame-gate caps-dark/blown/flat/content_small, graded must-acceptance), `asset_v1`, `reference_v1`, `shader_v2` (graphics, `gl_frames`+dark/blown/flat/content_small, graded must-acceptance), `asset`, `reference`, `shader` (graphics, `gl_frames`caps; also the `LikenessJudge` default). Retired `scene_v1` and `shader_v1` are no longer distributed.**Calibration** (`addons/calibration.py`: re-judges recorded rounds, n samples each) is why the@@ -485,7 +485,7 @@ runs at all; `3dcode resume --force` is the one way to regenerate a stage underthe round journal + `record.json` to `rounds/pre_force/` whether or not the spec changed, and the runstarts again at round 0 with no stage frozen. The plan itself is `run()`'s, before the graph (`plan_stage_key`). What thescene stages do: `assets` — both kinds climb one ladder (single-shot → check → one repair → agent;-blender_glb heroes get a sub-workspace with the static planner's parts + asset_v1 judge + one+blender_glb heroes get a sub-workspace with the static planner's parts + asset judge + onere-judged fix pass, undone when worse; a degraded asset verdict leaves score None / judged False,emits asset.judge_degraded and skips the fix pass); `env` — `env_file` (`src/env.js` / `src/env.py`): itssession shares the tree with the asset single-shots, so it is a `shared_tree` job (credited with its own filediff --git a/harness/docs/INTERFACES.md b/harness/docs/INTERFACES.mdindex 5f510845..2fe8119c 100644--- a/harness/docs/INTERFACES.md+++ b/harness/docs/INTERFACES.md@@ -416,7 +416,7 @@ render_blender_scene(ws, out_dir, *, cameras=None, orbit=True, times=(0.0, 1.5),# fallback, frame_ms, timings_ms}}; Settings.render.blender_{engine,samples,device,gpu_slots,gpu_wait_s}from codeverse3d.spatial.render_scene import select_judge_views, JUDGE_MAX_VIEWS, plan_boundsselect_judge_views(rs, max_n=10) -> RenderSet # priority: authored@t0, MAX_JUDGE_OVERVIEWS=4 overview@t0 (Δ 2026-10-14,- # scene_v2: the fourth is conventions.SCENE_VIEWS' overview_low_front, az 20° el 7°, orbit.mjs fitContactCamera), 2 authored@t_last, rest+ # scene: the fourth is conventions.SCENE_VIEWS' overview_low_front, az 20° el 7°, orbit.mjs fitContactCamera), 2 authored@t_last, restfrom codeverse3d.spatial.frame_metrics import frame_gate_from_renders, frame_findings, frame_summary_text, FRAME_GATEframe_gate_from_renders(renders) -> GateReport # gate "scene_frames"; data.kind ∈ dark_frame | blown_frame | flat_frame |# camera_in_geometry | camera_underground | camera_low | camera_high | content_small; authored cameras → ERROR, orbit rig → WARN@@ -470,7 +470,7 @@ placement_findings(table, *, indoor=False, language=None, thresholds=None) -> Ga# fog_short | world_edge (WARN; was no_backdrop ERROR: standing geometry ≥ WORLD_EDGE_MIN_HEIGHT_M past WORLD_EDGE_REACH_FACTOR ×# the half-extent) | density_note (INFO; was underdressed ERROR) | out_of_bounds (WARN on rows carrying a PLANNED content's name# only, margin max(BOUNDS_MARGIN_MIN_M 5, BOUNDS_MARGIN_FRAC 10 % of span); unplanned geometry past the bounds is never reported)- # messages carry the scene_v2 cap words (floating / sunken / interpenetration / scale / free_exempt); reads artifacts/census.json["placement"] (host_placement.mjs)+ # messages carry the scene cap words (floating / sunken / interpenetration / scale / free_exempt); reads artifacts/census.json["placement"] (host_placement.mjs)# Δ D105 (2026-10-14): thresholds = Settings.placement (config.Placement) read at CALL time, `thresholds=` overrides; the# summary INFO's data carries `thresholds` (what graded) + `counts` (measured) / `listed` (after the per-kind cap);# + kinds cantilevered (WARN: rests on ≤ 1 of ≥ 4 foot columns, host_placement rows gap_max_m / columns_resting / tilt_m)@@ -521,7 +521,7 @@ python -m codeverse3d.spatial.mcp_server --workspace <ws> [--track X] [--languag```pythonfrom codeverse3d.judges.rubrics import load_rubric, Rubric # Rubric{…, defects: [DefectItem{id, text, penalty, cap}], caps[{…, when:, kinds}]}from codeverse3d.judges.vlm_judge import VlmJudge-VlmJudge(rubric="static_object_v1", model_id=None (settings.default_judge = gemini-3.1-pro-preview), n_samples=1,+VlmJudge(rubric="static_object", model_id=None (settings.default_judge = gemini-3.1-pro-preview), n_samples=1,temperature=0.2, *, thinking=None, max_attempts=3, max_montages=None, detail_crops=None, max_px=None,sample_budget_s=None, fixed_order=False, chat_model=None, cache_dir=None, label="judge")# Δ max_images is GONE → montage budget. The None params resolve through Settings.judge@@ -546,7 +546,7 @@ plan_summary(plan, language) -> str # the digest for every track (object: "Ove# also the texture gate's (texturing/generate.judge_gate)from codeverse3d.judges.prompt_builder import plan_montages, render_montage, Montage # ≤5 2×2 montages (shaded/geometry/poses) + ≤2 detail# crops @≤1024px replace the sheet + the 14-view rig (D47); clay/normals views (RenderView.mode) auto-route to the GEOMETRY montage- # Δ 2026-10-14 (scene_v2): plan_montages(…, gates=()) / plan_detail_crops(shaded, *, scene, gates=(), limit) — a scene sends the+ # Δ 2026-10-14 (scene): plan_montages(…, gates=()) / plan_detail_crops(shaded, *, scene, gates=(), limit) — a scene sends the# centre crop + ≤ limit CONTACT CROPs (plan_contact_crops: a scene_placement finding's data.bbox projected through the judged# overview tile that shows it largest, project_box); gates_section(gates, *, scene=False) has a scene branch (placement →# frames → lint, WARNs grouped by kind, one MEASURED PLACEMENT line: placement_block); the object text is unchanged@@ -727,7 +727,7 @@ post-render gates → the build's own `BuildResult.gates` (also when the build f(when references), articulated `joint_sweep` + `motion_direction`, scene`render_console` + `scene_frames` (via `ctx.services.frame_gate` ->`frame_gate_from_renders`; advisory, a failure is logged not raised), graphics `gl_frames`. Reference specs get a-`ReferenceJudge` (rubric `reference_v1` for static_object; `LikenessJudge` on graphics/scene —+`ReferenceJudge` (rubric `reference` for static_object; `LikenessJudge` on graphics/scene —`vlm_judge.judge_for`) and images attached to generation prompts.## texturing/ (derived asset pack; code stays truth — object.glb is never touched)diff --git a/harness/docs/RUNBOOK.md b/harness/docs/RUNBOOK.mdindex 365947f3..558b8525 100644--- a/harness/docs/RUNBOOK.md+++ b/harness/docs/RUNBOOK.md@@ -181,7 +181,7 @@ unless you type `--host` yourself; it never serves a path outside the declared r# paid verdict per pair, cached); --round N = that round; --texture = the# texture pass on it first. `make`/`resume` run `pick --by score` for you3dcode render <slug> [--round N] [--mode shaded|wire|normals|clay|silhouette] [--out dir]-3dcode judge <slug> [--round N] [--rubric static_object_v1] [--model gemini:gemini-3.1-pro-preview] [--n 3] [--thinking off|low|medium|high]+3dcode judge <slug> [--round N] [--rubric static_object] [--model gemini:gemini-3.1-pro-preview] [--n 3] [--thinking off|low|medium|high]# re-judges a round's recorded renders → artifacts/judge/rNN_cli.json3dcode texture pass <slug> [--no-judge] [--model …] [--judge-model …] [--image-model …] [--size 1024]3dcode texture scene-pack <slug> [--n 10] · 3dcode texture show <slug>diff --git a/harness/runtime_js/lib/orbit.mjs b/harness/runtime_js/lib/orbit.mjsindex 599845ae..ff191a89 100644--- a/harness/runtime_js/lib/orbit.mjs+++ b/harness/runtime_js/lib/orbit.mjs@@ -137,7 +137,7 @@ const CONTACT_SPAN_FACTOR = 1.2; // eye distance from the content centre, in cconst CONTACT_EYE_M = 0.4; // eye height above the ground: a base's daylight is a few cm tall/**- * Ground-level CONTACT camera on the given azimuth side (scene_v2's `overview_low_front`,+ * Ground-level CONTACT camera on the given azimuth side (scene's `overview_low_front`,* 2026-10-14): the eye stands `CONTACT_SPAN_FACTOR` content spans from the centre, 0.4 m* above the ground, and looks at the point where the view line leaves the content box's* footprint AT ground height — so the near edge of the content sits mid-frame with thediff --git a/harness/tests/cost/conftest.py b/harness/tests/cost/conftest.pyindex 0142ee5f..55147619 100644--- a/harness/tests/cost/conftest.py+++ b/harness/tests/cost/conftest.py@@ -25,9 +25,9 @@ def fake_run(tmp_path: Path) -> Path:output_tokens=300, latency_ms=30_000), run=ws.name, round=0, stage="baseline",label="baseline", source="session", n_calls=2, ledger=ledger)record_call(Usage(backend="gemini", model="gemini-3.1-pro-preview", input_tokens=12_000, output_tokens=1_000,- latency_ms=5_000), run=ws.name, label="judge:static_object_v1:r00:s0", ledger=ledger)+ latency_ms=5_000), run=ws.name, label="judge:static_object:r00:s0", ledger=ledger)spec = Spec(id="fake", track=Track.STATIC_OBJECT, language=Language.BLENDER, prompt="a chair")- judgment = Judgment(rubric="static_object_v1", scores={}, overall=0.8, passed=True)+ judgment = Judgment(rubric="static_object", scores={}, overall=0.8, passed=True)rec = RunRecord(spec=spec, workspace=str(ws), status=RunStatus.MAX_ROUNDS, extra={"stop_reason": "max_rounds"},steps=[StepTime(step="plan", wall_s=20.0)],rounds=[RoundRecord(index=0, kind="baseline", judgment=judgment,diff --git a/harness/tests/cost/test_context.py b/harness/tests/cost/test_context.pyindex 042bf020..f7ea3663 100644--- a/harness/tests/cost/test_context.py+++ b/harness/tests/cost/test_context.py@@ -17,7 +17,7 @@ def test_label_tells_stage_role_and_round():c = context_from_label("refine_drip_tray")assert c.stage is Stage.REFINE and c.role is None # a generation label states no role...assert attribute(label="refine_drip_tray").role is Role.GENERATOR # ...the stage derives it- c = context_from_label("judge:static_object_v1:r02:s1")+ c = context_from_label("judge:static_object:r02:s1")assert c.stage is Stage.JUDGE and c.role is Role.JUDGE and c.round == 2assert context_from_label("planner-retry").stage is Stage.PLANassert context_from_label("pairwise:0v1:fwd").stage is Stage.PAIRWISE@@ -44,7 +44,7 @@ def test_session_beats_a_generation_label_a_named_job_beats_the_session_explicitassert tex.stage is Stage.TEXTURE and tex.role is Role.OTHER and tex.round == 2gate = attribute(label="texture_gate")assert gate.stage is Stage.TEXTURE and gate.role is Role.JUDGE- verdict = attribute(label="judge:static_object_v1:r02:s0")+ verdict = attribute(label="judge:static_object:r02:s0")assert verdict.stage is Stage.JUDGE and verdict.role is Role.JUDGE and verdict.round == 2cap = attribute(label="caption:best")assert cap.stage is Stage.CAPTION and cap.role is Role.CAPTIONERdiff --git a/harness/tests/cost/test_instrument.py b/harness/tests/cost/test_instrument.pyindex e6be7193..520a2e38 100644--- a/harness/tests/cost/test_instrument.py+++ b/harness/tests/cost/test_instrument.py@@ -83,7 +83,7 @@ def test_a_metered_call_becomes_one_priced_row(tmp_path: Path):chat = FakeChat()with run_ledger(tmp_path, run="r1"):resp = MeteredChatModel(chat).generate(- ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object_v1:r01:s0"))+ ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object:r01:s0"))assert resp.text == "ok"(row,) = load_ledger(tmp_path)assert row.run == "r1" and row.stage is Stage.JUDGE and row.role is Role.JUDGE and row.round == 1@@ -131,7 +131,7 @@ def test_post_hoc_work_never_creates_a_partial_ledger(tmp_path: Path):with run_ledger(tmp_path, run="old", create=False) as led:assert led is NoneMeteredChatModel(FakeChat()).generate(- ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object_v1:r00:s0"))+ ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object:r00:s0"))assert not (tmp_path / "telemetry" / "cost.jsonl").exists()# but a run that already keeps one is appended towith run_ledger(tmp_path, run="old"):@@ -139,7 +139,7 @@ def test_post_hoc_work_never_creates_a_partial_ledger(tmp_path: Path):with run_ledger(tmp_path, run="old", create=False) as led:assert led is not NoneMeteredChatModel(FakeChat()).generate(- ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object_v1:r00:s0"))+ ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object:r00:s0"))assert len(load_ledger(tmp_path)) == 2diff --git a/harness/tests/cost/test_live.py b/harness/tests/cost/test_live.pyindex 61fa66d1..3723bc08 100644--- a/harness/tests/cost/test_live.py+++ b/harness/tests/cost/test_live.py@@ -22,7 +22,7 @@ def _stable_text() -> str:from codeverse3d.judges.rubrics import load_rubricreturn "\n".join([- build_system_prompt(load_rubric("static_object_v1")),+ build_system_prompt(load_rubric("static_object")),(REPO / "codeverse3d/prompts/blender/contract.md").read_text(),(REPO / "codeverse3d/prompts/blender/system.md").read_text(),(REPO / "codeverse3d/prompts/system/harness_contract.md").read_text(),diff --git a/harness/tests/cost/test_tally.py b/harness/tests/cost/test_tally.pyindex 582da5df..391f91bd 100644--- a/harness/tests/cost/test_tally.py+++ b/harness/tests/cost/test_tally.py@@ -17,7 +17,7 @@ from codeverse3d.cost.instrument import MeteredAgent, MeteredChatModel, run_ledgfrom codeverse3d.cost.ledger import load_ledger, record_callfrom codeverse3d.cost.tally import Tally, tally, timed-REQ = ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object_v1:r00:s0")+REQ = ChatRequest(messages=[ChatMessage.user("x")], label="judge:static_object:r00:s0")def test_a_tally_holds_exactly_the_ledger_rows_booked_inside_it(tmp_path: Path):diff --git a/harness/tests/flywheel_cli/conftest.py b/harness/tests/flywheel_cli/conftest.pyindex a1bbf662..c3382d11 100644--- a/harness/tests/flywheel_cli/conftest.py+++ b/harness/tests/flywheel_cli/conftest.py@@ -45,7 +45,7 @@ def tiny_png(path: Path, rgb: tuple[int, int, int] = (200, 30, 30), size: int =def _judgment(score: float, passed: bool, plan: list[str]) -> Judgment:return Judgment(- rubric="static_object_v1", judge_backend="gemini:gemini-3.7-flash",+ rubric="static_object", judge_backend="gemini:gemini-3.7-flash",scores={"fidelity": score, "geometry": score}, overall=score, passed=passed, summary=f"score {score}",issues=[JudgeIssue(target="Leg", kind="geometry", severity="major", detail="legs too thin")] if not passed else [],improvement_plan=[ImprovementItem(target="Leg", kind="geometry", instruction=s, priority=1) for s in plan],diff --git a/harness/tests/flywheel_cli/test_cli.py b/harness/tests/flywheel_cli/test_cli.pyindex 2b6c607f..78581cfd 100644--- a/harness/tests/flywheel_cli/test_cli.py+++ b/harness/tests/flywheel_cli/test_cli.py@@ -165,7 +165,7 @@ def test_commands_over_a_runs_dir(runs_dir: Path, tmp_path: Path, monkeypatch):r = runner.invoke(app, ["judge", "wooden_chair_ab12cd34", "--runs-dir", str(runs_dir)])assert r.exit_code == 0, r.outputinp = _FakeVlm.captured["inp"]- assert _FakeVlm.captured["rubric"] == "static_object_v1" # the rubric the round was judged with+ assert _FakeVlm.captured["rubric"] == "static_object" # the rubric the round was judged withassert inp.round_index == 1 # best roundassert "Chair" in inp.plan_summary # plan digest, not emptyassert [a.id for a in inp.acceptance] == ["a1"]@@ -181,9 +181,9 @@ def test_commands_over_a_runs_dir(runs_dir: Path, tmp_path: Path, monkeypatch):monkeypatch.setattr(vj, "ReferenceJudge", _FakeRef)_FakeRef.captured.clear()- r = runner.invoke(app, ["judge", "wooden_chair_ab12cd34", "--runs-dir", str(runs_dir), "--rubric", "reference_v1"])+ r = runner.invoke(app, ["judge", "wooden_chair_ab12cd34", "--runs-dir", str(runs_dir), "--rubric", "reference"])assert r.exit_code == 0, r.output # no ValueError traceback any more- assert _FakeRef.captured["rubric"] == "reference_v1"+ assert _FakeRef.captured["rubric"] == "reference"# a judge ValueError is a CLI errorclass _Boom(_FakeVlm):@@ -213,7 +213,7 @@ def test_commands_over_a_runs_dir(runs_dir: Path, tmp_path: Path, monkeypatch):def test_calibration_rubric_map_includes_graphics(tmp_path: Path):- """A graphics round with no stored judgment is re-judged with the track's rubric, not static_object_v1."""+ """A graphics round with no stored judgment is re-judged with the track's rubric, not static_object."""from codeverse3d.addons.calibration import load_run_casesfrom codeverse3d.contracts.artifacts import RenderSet, RenderViewfrom codeverse3d.contracts.common import Language, Track@@ -226,7 +226,7 @@ def test_calibration_rubric_map_includes_graphics(tmp_path: Path):renders = RenderSet(views=[RenderView(name="frame_0", path=str(run / "artifacts" / "renders" / "r00" / "frame_0.png"))],renderer="fake")(run / "rounds" / "r00.json").write_text(RoundRecord(index=0, kind="baseline", renders=renders).model_dump_json())- assert [c.rubric for c in load_run_cases(run)] == ["shader_v2"]+ assert [c.rubric for c in load_run_cases(run)] == ["shader"]def test_make_invalid_combo_leaves_no_orphan_workspace(tmp_path: Path):@@ -657,7 +657,7 @@ def test_a_replay_judges_the_picked_round_on_the_runs_rubric_and_books_into_the_"""`3dcode judge` replays the run it reads. r0 0.55, r1 unjudged (a judge outage), r2 0.90, of astatic `--image` run:* no --round after `pick --round 0` is r0, the packaged round — not the best-scored r2 (N55a);- * the unjudged r1 replays on the run's reference_v1, not the track's static_object_v1 (N49);+ * the unjudged r1 replays on the run's reference, not the track's static_object (N49);* r2's previous verdict is the loop's — r1's, i.e. none — not r0's (N56);* what the replay (or a pick) paid joins record.total_usage: it stays the ledger's sum (N53)."""import codeverse3d.judges.vlm_judge as vj@@ -673,7 +673,7 @@ def test_a_replay_judges_the_picked_round_on_the_runs_rubric_and_books_into_the_r2 = rec.rounds[1].model_copy(update={"index": 2, "judgment": _judgment(0.90, True, [])})rec.rounds[1].judgment = Nonerec.rounds.append(r2)- rec.extra["rubric"] = "reference_v1"+ rec.extra["rubric"] = "reference"ws.write_json(ws.record_path, rec)ws.write_json(ws.spec_path, rec.spec)ledger = ws.telemetry / "cost.jsonl"@@ -703,7 +703,7 @@ def test_a_replay_judges_the_picked_round_on_the_runs_rubric_and_books_into_the_assert r.exit_code == 0, r.output(i0, _, _), (i1, rubric1, _), (i2, _, prev2) = _Judge.seenassert (i0, i1, i2) == (0, 1, 2)- assert rubric1 == "reference_v1"+ assert rubric1 == "reference"assert prev2 is Noneassert load_record(ws).total_usage.cost_usd == pytest.approx(ledger_usage(load_ledger(ws.root)).cost_usd)assert load_record(ws).total_usage.cost_usd == pytest.approx(total + 3 * 0.05)diff --git a/harness/tests/flywheel_cli/test_complexity_record.py b/harness/tests/flywheel_cli/test_complexity_record.pyindex 2da037f9..8c5e211b 100644--- a/harness/tests/flywheel_cli/test_complexity_record.py+++ b/harness/tests/flywheel_cli/test_complexity_record.py@@ -34,7 +34,7 @@ def _round(index: int, score: float, cx: float | None) -> RoundRecord:return RoundRecord(index=index, kind="baseline" if index == 0 else "refine", commit="c" * 12,measurement=_measurement(cx) if cx is not None else None,- judgment=Judgment(rubric="static_object_v1", judge_backend="gemini:x",+ judgment=Judgment(rubric="static_object", judge_backend="gemini:x",scores={"geometry_detail": score}, overall=score, passed=score >= 0.72,summary="s"),usage=Usage(cost_usd=0.5),diff --git a/harness/tests/flywheel_cli/test_run_identity.py b/harness/tests/flywheel_cli/test_run_identity.pyindex 2c7e744e..f3e4d4f3 100644--- a/harness/tests/flywheel_cli/test_run_identity.py+++ b/harness/tests/flywheel_cli/test_run_identity.py@@ -19,7 +19,7 @@ def test_the_slug_rule_and_the_battery_layouts():for battery, rel, slug in [("runs", "wooden_chair_ab12cd34", "wooden_chair_ab12cd34"),("h2h_scene_v1", "h2h_b01_neon_alley.attempt1", "h2h_b01_neon_alley.attempt1"), # a retry keeps its suffix- ("articulated_v1_flash", "runs/art_easy_laptop", "art_easy_laptop"), # runs/ is noise+ ("articulated_flash", "runs/art_easy_laptop", "art_easy_laptop"), # runs/ is noise("compare_v1_full", "cells/cmp_easy_stool/harness_x/run", "cmp_easy_stool__harness_x"),]:assert RunId(battery=battery, rel=rel).slug == slug, reldiff --git a/harness/tests/flywheel_cli/test_run_layout_cli.py b/harness/tests/flywheel_cli/test_run_layout_cli.pyindex 46efa280..cf3102da 100644--- a/harness/tests/flywheel_cli/test_run_layout_cli.py+++ b/harness/tests/flywheel_cli/test_run_layout_cli.py@@ -189,7 +189,7 @@ def test_show_sections_and_status_share_the_packaged_run(fake_run):assert "DELIVERABLE" in out and "QUALITY EVIDENCE" in out and "COST & SETTINGS" in outassert out.index("DELIVERABLE") < out.index("QUALITY EVIDENCE") < out.index("COST & SETTINGS")assert "deliverable/object.glb" in out and "deliverable/src/model.py" in out- assert "static_object_v1" in out # evidence: rubric+ assert "static_object" in out # evidence: rubricassert "the ledger, list price" in out and "models per role" in out and "price table" in outassert "gemini-cli" in out and "generator" in out # settings: model id per rolediff --git a/harness/tests/flywheel_cli/test_select.py b/harness/tests/flywheel_cli/test_select.pyindex 987ccec7..7f5b3c7c 100644--- a/harness/tests/flywheel_cli/test_select.py+++ b/harness/tests/flywheel_cli/test_select.py@@ -47,10 +47,10 @@ def _run(root: Path, rounds: list[tuple[float | str | None, int]], *, status: Rukept.parent.mkdir(parents=True, exist_ok=True)kept.write_bytes(b"glTF\x02\x00\x00\x00" + bytes([i]) * 16)if score == "degraded": # a judge glitch: paid for, never a score- j = Judgment(rubric="static_object_v1", scores={}, overall=0.0, passed=False,+ j = Judgment(rubric="static_object", scores={}, overall=0.0, passed=False,summary="judge_error: 503", raw=json.dumps({"status": "degraded"}))else:- j = None if score is None else Judgment(rubric="static_object_v1", scores={}, overall=score, passed=score >= 0.72)+ j = None if score is None else Judgment(rubric="static_object", scores={}, overall=score, passed=score >= 0.72)gates = [GateReport(gate="contract", passed=not errors,findings=[GateFinding(gate="contract", severity=Severity.ERROR, message="off")] * errors)]recs.append(RoundRecord(index=i, kind="baseline" if i == 0 else "refine", commit=commit, judgment=j, gates=gates,@@ -58,7 +58,7 @@ def _run(root: Path, rounds: list[tuple[float | str | None, int]], *, status: Rurenders=RenderSet(views=views, contact_sheet=str(tiny_png(ws.renders_dir(i) / "sheet.png"))),usage=Usage(cost_usd=0.1 * (i + 1))))ws.write_json(ws.record_path, RunRecord(spec=spec, workspace=str(ws.root), status=status, rounds=recs,- extra={"stop_reason": status.value, "rubric": "static_object_v1"}))+ extra={"stop_reason": status.value, "rubric": "static_object"}))return ws@@ -141,7 +141,7 @@ def test_pick_by_pairwise_asks_only_inside_the_margin_and_never_buys_a_verdict_tws = _run(tmp_path / "r", [(0.70, 0), (0.72, 0)])judge = _Pairwise("b", 0.9)assert select.pick(ws.root, by="pairwise", judge=judge) == 0 # r00 is B, and B won- assert judge.calls and "/r01/" in judge.calls[0][0] and judge.calls[0][2] == "static_object_v1"+ assert judge.calls and "/r01/" in judge.calls[0][0] and judge.calls[0][2] == "static_object"assert select.pick(ws.root, by="pairwise", judge=judge) == 0 and len(judge.calls) == 1 # cached, not re-boughtassert ws.judge_path(1, "_vs_r00_pairwise").is_file()Path(judge.calls[0][0]).write_bytes(b"re-rendered") # new pictures are a new comparisondiff --git a/harness/tests/graphics/test_live.py b/harness/tests/graphics/test_live.pyindex 8aedf855..3918afa8 100644--- a/harness/tests/graphics/test_live.py+++ b/harness/tests/graphics/test_live.py@@ -59,7 +59,7 @@ def test_live_plan_generate_build_judge(tmp_path, keys_ok):pytest.skip(f"model wrote a shader that does not compile (repair loop not exercised here): {build.error_message[:200]}")renders = frames_render_set(ws, build, 0)assert len(renders.views) == 5 and renders.contact_sheet- judge = VlmJudge(rubric="shader_v2", model_id=MODEL, n_samples=1)+ judge = VlmJudge(rubric="shader", model_id=MODEL, n_samples=1)j = judge.judge(JudgeInput(spec=spec, renders=renders, gates=[lint], acceptance=plan.acceptance, plan_summary=plan.summary,round_index=0, extra_context=""))assert 0.0 <= j.overall <= 1.0 and set(j.scores) >= {"brief_fidelity", "motion_quality"}diff --git a/harness/tests/graphics/test_prompt_numbers.py b/harness/tests/graphics/test_prompt_numbers.pyindex 9d255bf2..8c11e00e 100644--- a/harness/tests/graphics/test_prompt_numbers.py+++ b/harness/tests/graphics/test_prompt_numbers.py@@ -1,5 +1,5 @@"""The judge numbers ``generate_graphics.j2`` states are the rubric's (the gate thresholds it states are-pinned in ``prompts/_claims.toml``): change ``shader_v2.yaml`` and this fails until the prompt agrees."""+pinned in ``prompts/_claims.toml``): change ``shader.yaml`` and this fails until the prompt agrees."""from __future__ import annotations@@ -9,7 +9,7 @@ from codeverse3d.prompts import load_textdef test_generate_graphics_states_the_rubrics_own_numbers() -> None:text = load_text("tracks/generate_graphics.j2")- rubric = load_rubric("shader_v2")+ rubric = load_rubric("shader")caps = {c.id: c.cap for c in rubric.caps}penalties = {d.id: d.penalty for d in rubric.defects}depth = next(c for c in rubric.criteria if c.id == "depth_and_detail")diff --git a/harness/tests/judges/test_calibration.py b/harness/tests/judges/test_calibration.pyindex f5ce3f90..3f8ac808 100644--- a/harness/tests/judges/test_calibration.py+++ b/harness/tests/judges/test_calibration.py@@ -16,7 +16,7 @@ from tests.judges.conftest import (make_spec,)-R = load_rubric("static_object_v1")+R = load_rubric("static_object")def _fake_run(root: Path, name: str) -> Path:@@ -51,7 +51,7 @@ def test_load_cases_reads_the_plan_and_picks_the_fallback_round(tmp_path, caplogrun = _fake_run(tmp_path, "runA")cases = load_run_cases(run)assert [c.round_index for c in cases] == [0, 1] and cases[0].gate_errors == 1 and cases[1].gate_errors == 0- assert cases[0].rubric == "static_object_v1" and len(cases[0].inp.acceptance) == 2+ assert cases[0].rubric == "static_object" and len(cases[0].inp.acceptance) == 2assert cases[1].is_picked and not cases[0].is_picked and cases[0].glb is None # no usable record: the last roundassert cases[0].inp.plan_summary == "Chair: a chair. Overall 0.50×0.90×0.50 m (W×H×D). Parts: Seat, Leg×4."assert load_run_cases(run, rounds=[1])[0].round_index == 1diff --git a/harness/tests/judges/test_caps.py b/harness/tests/judges/test_caps.pyindex 4524c2db..5168b464 100644--- a/harness/tests/judges/test_caps.py+++ b/harness/tests/judges/test_caps.py@@ -12,9 +12,9 @@ from codeverse3d.judges.rubrics import (measured_absent,)-R = load_rubric("static_object_v1")-A = load_rubric("articulated_v1")-S = load_rubric("scene_v2")+R = load_rubric("static_object")+A = load_rubric("articulated")+S = load_rubric("scene")POSE = [RenderView(name="pose_rest", path="p")]@@ -50,7 +50,7 @@ E, W = Severity.ERROR, Severity.WARN{"views": POSE}, 0.6, {"floating_part"}),(S, [], {"console_errors": ["TypeError: x is undefined"]}, 0.6, {"console_error"}),(S, [_gate("shader_probe", E, "GLSL compile failed in water.js", kind="shader")], {}, 0.5, {"shader_error"}),- # scene_v2: ONE cap per placement kind, by the finding's typed kind (the v1 gate-'*' floating_part+ # scene: ONE cap per placement kind, by the finding's typed kind (the v1 gate-'*' floating_part# twin that fired beside floating_or_sunken_asset is gone — a floating ERROR is a single 0.60 cap)(S, [_gate("scene_placement", E, "Pondside/Lantern is floating 0.30 m above PondWater", target="Pondside/Lantern",kind="floating", gap_m=0.30)], {"defects_present": {}}, 0.6, {"floating_asset"}),@@ -80,7 +80,7 @@ def test_gate_caps(rubric, gates, kw, overall, rules):assert applied == rules if rules == set() else rules <= applied-def test_scene_v2_placement_caps_fire_one_each_and_the_tightest_wins():+def test_scene_placement_caps_fire_one_each_and_the_tightest_wins():"""Three measured placement ERRORs = three ledger lines (one per kind) and the 0.60 floor."""gates = [GateReport(gate="scene_placement", passed=False, findings=[GateFinding(gate="scene_placement", severity=E, target="A", message="A is floating 0.12 m", data={"kind": "floating", "gap_m": 0.12}),@@ -164,11 +164,11 @@ def _placement(passed: bool, *findings: tuple[Severity, str, str, dict]) -> Gate# the gate did not run("floating_asset", [GateReport(gate="scene_frames", passed=True)], False),])-def test_scene_v2_veto_is_scoped_to_the_rules_kinds(defect, gates, absent):+def test_scene_veto_is_scoped_to_the_rules_kinds(defect, gates, absent):assert measured_absent(S, defect, gates) is absent-def test_scene_v2_veto_rides_through_aggregate_samples():+def test_scene_veto_rides_through_aggregate_samples():"""The vote marks floating_asset; the passed placement gate switches it off (no penalty, no cap),and a no_fog ERROR in the same report changes nothing about that."""sample = JudgeOutput(criteria={c.id: CriterionScore(score=0.9, evidence="") for c in S.criteria},@@ -190,7 +190,7 @@ def test_graded_acceptance_cap_keeps_the_gradient_but_not_the_pass():assert apply_caps(R, 0.99, [], {"M0": True}, ten[:2]).overall == pytest.approx(0.8)should = [AcceptanceItem(id="S1", text="y", priority="should")]assert apply_caps(R, 0.9, [], {"S1": False}, should).overall == 0.9, "a should item never caps"- assert apply_caps(S, 0.99, [], nine_ok, ten).overall == pytest.approx(0.96), "scene_v2 grades it like the object rubrics"+ assert apply_caps(S, 0.99, [], nine_ok, ten).overall == pytest.approx(0.96), "scene grades it like the object rubrics"# pass/fail is the owner's policy: the verdict still fails on must_missingsample = JudgeOutput(criteria={c.id: CriterionScore(score=0.99, evidence="") for c in R.criteria},acceptance={k: AcceptanceVerdict(verified=v) for k, v in nine_ok.items()},diff --git a/harness/tests/judges/test_defects.py b/harness/tests/judges/test_defects.pyindex 61cc90b4..44f14751 100644--- a/harness/tests/judges/test_defects.py+++ b/harness/tests/judges/test_defects.py@@ -18,8 +18,8 @@ from codeverse3d.judges.rubrics import (from codeverse3d.judges.vlm_judge import VlmJudgefrom tests.judges.conftest import FakeChatModel, good_reply-R = load_rubric("static_object_v1")-A = load_rubric("articulated_v1")+R = load_rubric("static_object")+A = load_rubric("articulated")IDS = ["A1", "A2"]@@ -30,7 +30,7 @@ def _reply(score=0.9, defects=None):def test_rubrics_declare_defects_with_costs():- for name in ("static_object_v1", "articulated_v1", "scene_v2", "asset_v1", "reference_v1"):+ for name in ("static_object", "articulated", "scene", "asset", "reference"):r = load_rubric(name)assert r.defects, nameassert all(d.penalty > 0 or d.cap is not None for d in r.defects)@@ -74,14 +74,14 @@ def test_wire_schema_has_fixed_defect_keys_and_parse_variants():def test_penalties_and_caps_computed_in_code(judge_input, cache_dir):# floating_part: -0.10 and cap 0.60; primitive_only: -0.08 no capmodel = FakeChatModel([_reply(0.9, ["primitive_only"])])- j = VlmJudge("static_object_v1", chat_model=model, cache_dir=cache_dir).judge(judge_input)+ j = VlmJudge("static_object", chat_model=model, cache_dir=cache_dir).judge(judge_input)raw = json.loads(j.raw)assert raw["overall_uncapped"] == pytest.approx(0.9) and raw["defect_penalty"] == pytest.approx(0.08)assert j.overall == pytest.approx(0.82) and j.passedassert raw["defects"]["primitive_only"] is True and raw["caps"]["caps_applied"] == []assert "defects (-0.08): primitive_only" in j.summarymodel2 = FakeChatModel([_reply(0.9, ["floating_part", "primitive_only"])])- j2 = VlmJudge("static_object_v1", chat_model=model2, cache_dir=cache_dir).judge(judge_input)+ j2 = VlmJudge("static_object", chat_model=model2, cache_dir=cache_dir).judge(judge_input)raw2 = json.loads(j2.raw)assert raw2["overall_after_defects"] == pytest.approx(0.72)assert j2.overall == 0.6 and not j2.passed # capped by defect:floating_part@@ -94,16 +94,16 @@ def test_defect_vote_ties_are_absent(judge_input, cache_dir, caplog):"""D36: an exact defect-vote tie is ABSENT, whichever sample flagged it."""flagged_first = FakeChatModel(by_label={":s0": [_reply(0.9, ["render_artifacts"])], ":s1": [_reply(0.9)]})with caplog.at_level("WARNING", logger="codeverse3d.judges.vlm_judge"):- j = VlmJudge("static_object_v1", chat_model=flagged_first, n_samples=2, cache_dir=cache_dir).judge(judge_input)+ j = VlmJudge("static_object", chat_model=flagged_first, n_samples=2, cache_dir=cache_dir).judge(judge_input)assert "n_samples=2 is even" in caplog.textraw = json.loads(j.raw)assert raw["defect_votes"]["render_artifacts"] == [True, False]assert raw["defects"]["render_artifacts"] is False and j.overall == pytest.approx(0.9)assert raw["tie_broken"] == ["render_artifacts"]- assert raw["scoring_version"] == 4 and raw["overridden"] == [] # 4: the kind-scoped veto (scene_v2, 2026-10-14)+ assert raw["scoring_version"] == 4 and raw["overridden"] == [] # 4: the kind-scoped veto (scene, 2026-10-14)clean_first = FakeChatModel(by_label={":s0": [_reply(0.9)], ":s1": [_reply(0.9, ["render_artifacts"])]})- j2 = VlmJudge("static_object_v1", chat_model=clean_first, n_samples=2, cache_dir=cache_dir).judge(judge_input)+ j2 = VlmJudge("static_object", chat_model=clean_first, n_samples=2, cache_dir=cache_dir).judge(judge_input)raw2 = json.loads(j2.raw)assert raw2["defect_votes"]["render_artifacts"] == [False, True]assert raw2["defects"]["render_artifacts"] is False and j2.overall == pytest.approx(0.9)@@ -114,12 +114,12 @@ def test_acceptance_vote_ties_follow_the_representative_sample(judge_input, cach"""D36: an acceptance tie follows the representative sample."""a_ok, a_no = _reply(0.9), _reply(0.9)a_no["acceptance"]["A1"]["verified"] = False- j = VlmJudge("static_object_v1", chat_model=FakeChatModel(by_label={":s0": [a_ok], ":s1": [a_no]}),+ j = VlmJudge("static_object", chat_model=FakeChatModel(by_label={":s0": [a_ok], ":s1": [a_no]}),n_samples=2, cache_dir=cache_dir).judge(judge_input)assert j.acceptance_results["A1"] is True and json.loads(j.raw)["tie_broken"] == ["A1"]b_ok, b_no = _reply(0.9), _reply(0.9)b_no["acceptance"]["A1"]["verified"] = False- j2 = VlmJudge("static_object_v1", chat_model=FakeChatModel(by_label={":s0": [b_no], ":s1": [b_ok]}),+ j2 = VlmJudge("static_object", chat_model=FakeChatModel(by_label={":s0": [b_no], ":s1": [b_ok]}),n_samples=2, cache_dir=cache_dir).judge(judge_input)assert j2.acceptance_results["A1"] is False and json.loads(j2.raw)["tie_broken"] == ["A1"]diff --git a/harness/tests/judges/test_detail_anchors.py b/harness/tests/judges/test_detail_anchors.pyindex ede3fd6d..8069c772 100644--- a/harness/tests/judges/test_detail_anchors.py+++ b/harness/tests/judges/test_detail_anchors.py@@ -9,8 +9,8 @@ from codeverse3d.judges.rubrics import _rule_hit, load_rubricdef test_detail_anchors_are_a_graded_countable_geometry_ladder() -> None:- """reference_v1 only: the other rubrics' system prompts are pinned by tests/prompts/manifest.json."""- r = load_rubric("reference_v1")+ """reference only: the other rubrics' system prompts are pinned by tests/prompts/manifest.json."""+ r = load_rubric("reference")c = r.criterion("detail")levels = sorted((float(k) for k in c.anchors), reverse=True)assert len(levels) >= 5 and levels[0] == 1.0 and levels[-1] == 0.1, levels@@ -23,18 +23,18 @@ def test_detail_anchors_are_a_graded_countable_geometry_ladder() -> None:def test_physical_plausibility_caps_are_untouched() -> None:"""The complexity wave may sharpen detail; it may never soften the gates."""- caps = {c.id: c for c in load_rubric("static_object_v1").caps}+ caps = {c.id: c for c in load_rubric("static_object").caps}assert caps["build_error"].cap == 0.0assert caps["floating_part"].cap == 0.6assert caps["penetration_error"].cap == 0.7assert caps["contract_violation"].cap == 0.75assert caps["missing_must_acceptance"].cap == 0.6- defects = {d.id: d for d in load_rubric("static_object_v1").defects}+ defects = {d.id: d for d in load_rubric("static_object").defects}assert defects["floating_part"].cap == 0.6 and defects["interpenetration"].cap == 0.7assert defects["wrong_object"].cap == 0.25[email protected]("rubric", ["static_object_v1", "articulated_v1", "scene_v2"])[email protected]("rubric", ["static_object", "articulated", "scene"])def test_acceptance_cap_is_graded_with_the_floor_pinned(rubric: str) -> None:"""D46 b: the acceptance cap is graded, 0.6 + 0.4·verified/total — the flag and the floor pinned."""rule = next(c for c in load_rubric(rubric).caps if c.id == "missing_must_acceptance")diff --git a/harness/tests/judges/test_judge_slices.py b/harness/tests/judges/test_judge_slices.pyindex 62434233..fe85a02e 100644--- a/harness/tests/judges/test_judge_slices.py+++ b/harness/tests/judges/test_judge_slices.py@@ -28,7 +28,7 @@ from codeverse3d.judges.vlm_judge import SLICE_LABELS, VlmJudgefrom codeverse3d.workspace import Workspacefrom tests.judges.conftest import ACCEPTANCE, FakeChatModel, good_reply, make_renders, make_spec-R = load_rubric("static_object_v1")+R = load_rubric("static_object")IDS = ["A1", "A2"]@@ -69,7 +69,7 @@ def make_input(renders: RenderSet, gates: list[GateReport], glb: Path | None, **def judge_request(inp: JudgeInput, cache_dir: Path):model = FakeChatModel([good_reply(R, IDS, 0.8)])- VlmJudge("static_object_v1", model_id="fake:fake-1", chat_model=model, cache_dir=cache_dir).judge(inp)+ VlmJudge("static_object", model_id="fake:fake-1", chat_model=model, cache_dir=cache_dir).judge(inp)return model.requests[0]@@ -119,7 +119,7 @@ def test_slices_ride_only_on_a_connectivity_error(tmp_path, cache_dir):assert PROVENANCE_ELICITATION not in req.systemassert "cross-section" not in text_of(req)# and the protocol hash is untouched: the hash builds the default (elicitation-free) prompt- assert judge_prompt_hash(R) == VlmJudge("static_object_v1", model_id="fake:fake-1").prompt_hash+ assert judge_prompt_hash(R) == VlmJudge("static_object", model_id="fake:fake-1").prompt_hashreq = judge_request(make_input(renders, dirty_gates(), glb), cache_dir)# placement: montage, 2 detail crops, THEN the two slices — never prependeddiff --git a/harness/tests/judges/test_likeness.py b/harness/tests/judges/test_likeness.pyindex 892786ec..9e2719fd 100644--- a/harness/tests/judges/test_likeness.py+++ b/harness/tests/judges/test_likeness.py@@ -14,8 +14,8 @@ def test_likeness_context_attaches_the_photos_and_the_note(judge_input, tmp_pathdraw_chair(ref)judge_input.spec = make_spec(track=Track.GRAPHICS, language=Language.GLSL_SHADER,references=[ReferenceImage(path=str(ref), role="likeness", note="aurora green spiral sea")])- model = FakeChatModel([good_reply(load_rubric("shader_v2"), ["A1", "A2"], 0.8)])- j = LikenessJudge("fake:fake-1", rubric="shader_v2", chat_model=model, cache_dir=cache_dir)+ model = FakeChatModel([good_reply(load_rubric("shader"), ["A1", "A2"], 0.8)])+ j = LikenessJudge("fake:fake-1", rubric="shader", chat_model=model, cache_dir=cache_dir)ctx = j.context(judge_input)assert ctx.extra_images == [("REAL-WORLD REFERENCE 1/1 — aurora green spiral sea", str(ref))]assert ctx.extra_text == LIKENESS_NOTE and "evenly spaced bars" in LIKENESS_NOTEdiff --git a/harness/tests/judges/test_live.py b/harness/tests/judges/test_live.pyindex ec02ff37..d9df86b0 100644--- a/harness/tests/judges/test_live.py+++ b/harness/tests/judges/test_live.py@@ -64,7 +64,7 @@ def test_live_vlm_judge_two_samples(tmp_path):acceptance=[AcceptanceItem(id="A1", text="three legs reach the ground", how="visual", priority="must"),AcceptanceItem(id="A2", text="seat is round", how="visual", priority="should")],plan_summary="parts: Seat, Leg1, Leg2, Leg3")- judge = VlmJudge("static_object_v1", model_id=MODEL, n_samples=2, cache_dir=tmp_path / "cache")+ judge = VlmJudge("static_object", model_id=MODEL, n_samples=2, cache_dir=tmp_path / "cache")j = judge.judge(inp)raw = json.loads(j.raw)print(f"\nLIVE source={source}\noverall={j.overall} std={j.score_std} passed={j.passed} n={j.n_samples}"@@ -84,7 +84,7 @@ def test_live_judge_separates_crafted_from_crude(tmp_path):if "stool" not in source:pytest.skip("needs real stool renders")spec = make_spec(prompt="A simple three-legged wooden stool, 45 cm tall, round seat 36 cm across.")- judge = VlmJudge("static_object_v1", model_id=MODEL, n_samples=1, cache_dir=tmp_path / "cache")+ judge = VlmJudge("static_object", model_id=MODEL, n_samples=1, cache_dir=tmp_path / "cache")good = judge.judge(JudgeInput(spec=spec, renders=renders, measurement=meas, plan_summary="parts: Seat, Leg1, Leg2, Leg3"))crude = judge.judge(JudgeInput(spec=spec, renders=make_renders(tmp_path / "crude"), measurement=make_measurement(),plan_summary="parts: Seat, Leg1, Leg2, Leg3"))diff --git a/harness/tests/judges/test_output_schema.py b/harness/tests/judges/test_output_schema.pyindex 3529e8f8..56a8d2f5 100644--- a/harness/tests/judges/test_output_schema.py+++ b/harness/tests/judges/test_output_schema.py@@ -8,7 +8,7 @@ from codeverse3d.judges.rubrics import ()from tests.judges.conftest import good_reply-R = load_rubric("static_object_v1")+R = load_rubric("static_object")IDS = ["A1", "must-2"]diff --git a/harness/tests/judges/test_prompt_builder.py b/harness/tests/judges/test_prompt_builder.pyindex 96f776df..992cab63 100644--- a/harness/tests/judges/test_prompt_builder.py+++ b/harness/tests/judges/test_prompt_builder.py@@ -20,7 +20,7 @@ from codeverse3d.judges.prompt_builder import (from codeverse3d.judges.rubrics import load_rubricfrom tests.judges.conftest import draw_chair-R = load_rubric("static_object_v1")+R = load_rubric("static_object")def _parts(msgs):@@ -64,7 +64,7 @@ def test_gates_previous_and_budget(judge_input, cache_dir):judge_input.gates = [GateReport(gate="connectivity", passed=False, findings=[GateFinding(gate="connectivity", severity=Severity.ERROR, target="LegBackLeft", message="floating 3 cm"),GateFinding(gate="connectivity", severity=Severity.WARN, message="thin sliver on Seat")])]- judge_input.previous = Judgment(rubric="static_object_v1", scores={}, overall=0.55, passed=False,+ judge_input.previous = Judgment(rubric="static_object", scores={}, overall=0.55, passed=False,improvement_plan=[ImprovementItem(target="Seat", kind="geometry", instruction="thicken seat", priority=1)])judge_input.round_index = 2judge_input.extra_context = "x" * 50_000@@ -80,10 +80,10 @@ def test_scene_rig_paragraph(judge_input, cache_dir):from codeverse3d.contracts.common import Language, Trackjudge_input.spec = judge_input.spec.model_copy(update={"track": Track.SCENE, "language": Language.SCENE_THREEJS})judge_input.renders.console_errors = ["ReferenceError: foo"]- _, msgs = build_judge_messages(judge_input, load_rubric("scene_v2"), cache_dir=cache_dir)+ _, msgs = build_judge_messages(judge_input, load_rubric("scene"), cache_dir=cache_dir)text = _parts(msgs)[0].textassert "overview_*" in text and "console error" in text- # scene_v2: the rig paragraph says what the overview tiles hide and what a CONTACT CROP is+ # scene: the rig paragraph says what the overview tiles hide and what a CONTACT CROP isassert "room ceilings lifted" in text and "CONTACT CROP images are diagnostic close-ups" in textassert "must name its view and its asset" in textassert len(_images(msgs)) == 2 # montage + centre crop only: no placement finding, no contact cropdiff --git a/harness/tests/judges/test_reference.py b/harness/tests/judges/test_reference.pyindex 5dda0338..4d0dd5a4 100644--- a/harness/tests/judges/test_reference.py+++ b/harness/tests/judges/test_reference.py@@ -5,7 +5,7 @@ from codeverse3d.judges.rubrics import load_rubricfrom codeverse3d.judges.vlm_judge import ReferenceJudge, ReferenceJudgeError, iou_to_scorefrom tests.judges.conftest import FakeChatModel, draw_chair, good_reply, image_parts-REF = load_rubric("reference_v1")+REF = load_rubric("reference")def test_reference_judge_measures_silhouette_and_adds_reference_images(judge_input, tmp_path, cache_dir):diff --git a/harness/tests/judges/test_rubrics.py b/harness/tests/judges/test_rubrics.pyindex 82274a0a..45ee76cf 100644--- a/harness/tests/judges/test_rubrics.py+++ b/harness/tests/judges/test_rubrics.py@@ -23,12 +23,12 @@ def test_cap_rules_are_not_part_of_the_judge_prompt():from codeverse3d.judges.prompt_builder import judge_prompt_hashfrom codeverse3d.judges.rubrics import RUBRICS_DIR, rubric_from_dict- r = load_rubric("static_object_v1")- data = yaml.safe_load((RUBRICS_DIR / "static_object_v1.yaml").read_text())+ r = load_rubric("static_object")+ data = yaml.safe_load((RUBRICS_DIR / "static_object.yaml").read_text())data["caps"] = []stripped = rubric_from_dict(data)assert stripped.content_hash() != r.content_hash(), "content_hash hashes what the YAML declares"# …and only that: a schema default the YAML never wrote does not re-key recorded verdicts- assert r.content_hash() == rubric_from_dict(yaml.safe_load((RUBRICS_DIR / "static_object_v1.yaml").read_text())).content_hash()+ assert r.content_hash() == rubric_from_dict(yaml.safe_load((RUBRICS_DIR / "static_object.yaml").read_text())).content_hash()assert judge_prompt_hash(stripped) == judge_prompt_hash(r)assert "interpenetration" in judge_prompt_hash.__globals__["build_system_prompt"](r), "the checklist defect IS rendered"diff --git a/harness/tests/judges/test_scene_judge.py b/harness/tests/judges/test_scene_judge.pyindex 4dee6f91..daaf5e18 100644--- a/harness/tests/judges/test_scene_judge.py+++ b/harness/tests/judges/test_scene_judge.py@@ -1,4 +1,4 @@-"""scene_v2 (2026-10-14): what the scene judge is shown — the gate facts in the scene branch, the+"""scene (2026-10-14): what the scene judge is shown — the gate facts in the scene branch, theMEASURED PLACEMENT line, the contact crops cut around measured placement findings, and the rubric'sown numbers. The object branch of every helper is pinned byte-identical in test_prompt_builder."""@@ -34,7 +34,7 @@ from tests.judges.conftest import draw_chair, make_specE, W, INFO = Severity.ERROR, Severity.WARN, Severity.INFOG = "scene_placement"-S = load_rubric("scene_v2")+S = load_rubric("scene")# --------------------------------------------------------------------------- fixtures@@ -273,7 +273,7 @@ def test_a_scene_verdict_sends_the_contact_crops_after_the_centre_crop(tmp_path,assert text.index("scene_placement=FAIL") < text.index("scene_frames=pass")# an object verdict never gets the scene textobj = JudgeInput(spec=make_spec(), renders=RenderSet(views=[authored]), gates=[_placement(LANTERN)])- _, omsgs = build_judge_messages(obj, load_rubric("static_object_v1"), cache_dir=cache_dir)+ _, omsgs = build_judge_messages(obj, load_rubric("static_object"), cache_dir=cache_dir)otext = omsgs[0].parts[0].textassert "MEASURED PLACEMENT" not in otext and "CONTACT CROP" not in otext and RIG_RULES["object"] in otext@@ -287,7 +287,7 @@ def test_view_rig_counts_contact_crops_as_detail_crops():# --------------------------------------------------------------------------- the rubric-def test_scene_v2_numbers_are_pinned():+def test_scene_numbers_are_pinned():"""What moved from v1 (eval/docs/EVAL.md §6, 2026-10-14) — and what stayed."""caps = {c.id: c for c in S.caps}assert S.version == 2 and S.pass_threshold == 0.70@@ -339,7 +339,7 @@ def test_scene_v2_numbers_are_pinned():"CONTACT CROP images are diagnostic close-ups", "must name the VIEW"):assert phrase in notes, phrase# the object rubrics did not move: no veto_warn_keys, the same content hash as their YAML- for name in ("static_object_v1", "articulated_v1", "reference_v1"):+ for name in ("static_object", "articulated", "reference"):assert all(not c.veto_warn_keys for c in load_rubric(name).caps), namediff --git a/harness/tests/judges/test_vlm_judge.py b/harness/tests/judges/test_vlm_judge.pyindex fa1ba27e..01f83387 100644--- a/harness/tests/judges/test_vlm_judge.py+++ b/harness/tests/judges/test_vlm_judge.py@@ -7,12 +7,12 @@ from codeverse3d.judges.vlm_judge import VlmJudgefrom codeverse3d.models.base import ModelErrorfrom tests.judges.conftest import FakeChatModel, good_reply, image_parts-R = load_rubric("static_object_v1")+R = load_rubric("static_object")IDS = ["A1", "A2"]def _judge(model, **kw):- return VlmJudge("static_object_v1", model_id="fake:fake-1", chat_model=model, **kw)+ return VlmJudge("static_object", model_id="fake:fake-1", chat_model=model, **kw)def test_full_path_single_sample(judge_input, cache_dir):@@ -23,13 +23,13 @@ def test_full_path_single_sample(judge_input, cache_dir):assert j.acceptance_results == {"A1": True, "A2": True}assert j.improvement_plan[0].instruction.startswith("bevel")assert j.usage.cost_usd == pytest.approx(0.001)- assert j.judge_backend == "fake:fake-1" and j.rubric == "static_object_v1"+ assert j.judge_backend == "fake:fake-1" and j.rubric == "static_object"raw = json.loads(j.raw)assert raw["status"] == "ok" and raw["n_used"] == 1 and raw["rubric_hash"] == R.content_hash()assert "[verdict: overall 0.80 vs threshold 0.72 → PASS]" in j.summaryreq = model.requests[0]assert req.response_schema is not None and "Criteria" in req.response_schema["$defs"]- assert req.label == "judge:static_object_v1:r00:s0" and req.temperature == 0.2+ assert req.label == "judge:static_object:r00:s0" and req.temperature == 0.2assert len(image_parts(req)) == 3 # 2×2 montage + 2 detail cropsassert "Defects" in req.response_schema["$defs"]assert raw["defects"] == {d.id: False for d in R.defects} and raw["defect_penalty"] == 0.0@@ -85,7 +85,7 @@ def test_a_failing_model_degrades_the_verdict(judge_input, cache_dir, retryable,def test_base_judge_rejects_measured_rubric(judge_input, cache_dir):with pytest.raises(ValueError, match="measured criteria"):- VlmJudge("reference_v1", chat_model=FakeChatModel(), cache_dir=cache_dir).judge(judge_input)+ VlmJudge("reference", chat_model=FakeChatModel(), cache_dir=cache_dir).judge(judge_input)# --------------------------------------------------------------- payload size (docs/COST.md §3)@@ -97,16 +97,16 @@ def test_judge_payload_size_comes_from_the_settings_dial(monkeypatch):s = Settings()s.apply_profile("economy", force=True)monkeypatch.setattr("codeverse3d.judges.vlm_judge.get_settings", lambda: s)- j = VlmJudge(rubric="static_object_v1", model_id="fake:fake-1")+ j = VlmJudge(rubric="static_object", model_id="fake:fake-1")# no profile shrinks the payload (docs/COST.md §14, D47)assert j.max_px == 1024 and j.detail_crops == 2 and j.max_montages == 5# a caller that states a size still gets it — the dial is the default, not a cap- explicit = VlmJudge(rubric="static_object_v1", model_id="fake:fake-1", max_px=768, detail_crops=1)+ explicit = VlmJudge(rubric="static_object", model_id="fake:fake-1", max_px=768, detail_crops=1)assert explicit.max_px == 768 and explicit.detail_crops == 1s2 = Settings(judge={"detail_crops": 0, "max_px": 512})monkeypatch.setattr("codeverse3d.judges.vlm_judge.get_settings", lambda: s2)s2.apply_profile("economy") # a stated payload survives a profile- assert VlmJudge(rubric="static_object_v1", model_id="fake:fake-1").detail_crops == 0+ assert VlmJudge(rubric="static_object", model_id="fake:fake-1").detail_crops == 0get_settings.cache_clear()@@ -121,7 +121,7 @@ def test_judge_prompt_hash_tracks_the_protocol_not_the_run(monkeypatch):# per-run content (acceptance ids) is NOT part of it: the schema is hashed with no idsassert wire_schema(R, ["A1"]) != wire_schema(R, []) and judge_prompt_hash(R) == base# the rubric is part of it (through the rendered rubric block)- other = load_rubric("shader_v2")+ other = load_rubric("shader")assert judge_prompt_hash(other) != base# an edit to the role prompt or the rig rules changes it — that is the whole pointmonkeypatch.setattr(pb, "_ROLE", pb._ROLE + "\nScore generously.")@@ -170,7 +170,7 @@ def test_a_sample_stops_when_its_budget_is_spent(judge_input, cache_dir, monkeypdef test_per_track_montage_ceilings():"""D47's 5 montages were measured on the object rig only: graphics keeps 3; the scene rig- takes 4 since scene_v2 (the fourth overview is the contact camera the crops are cut from)."""+ takes 4 since scene (the fourth overview is the contact camera the crops are cut from)."""from codeverse3d.contracts.artifacts import RenderSetfrom codeverse3d.contracts.common import Language, Trackfrom codeverse3d.contracts.spec import Spec@@ -182,13 +182,13 @@ def test_per_track_montage_ceilings():return JudgeInput(spec=spec, renders=RenderSet(), gates=[], acceptance=[])assert TRACK_MONTAGE_CAP == {"scene": 4, "graphics": 3}- j = VlmJudge("scene_v2", model_id="fake:fake-1")+ j = VlmJudge("scene", model_id="fake:fake-1")assert j._montages_for(inp(Track.SCENE, Language.SCENE_THREEJS)) == 4assert j._montages_for(inp(Track.SCENE, Language.SCENE_BLENDER)) == 4assert j._montages_for(inp(Track.GRAPHICS, Language.GLSL_SHADER)) == 3- jo = VlmJudge("static_object_v1", model_id="fake:fake-1")+ jo = VlmJudge("static_object", model_id="fake:fake-1")assert jo._montages_for(inp(Track.STATIC_OBJECT, Language.BLENDER)) == 5- explicit = VlmJudge("scene_v2", model_id="fake:fake-1", max_montages=5)+ explicit = VlmJudge("scene", model_id="fake:fake-1", max_montages=5)assert explicit._montages_for(inp(Track.SCENE, Language.SCENE_THREEJS)) == 5@@ -204,14 +204,14 @@ def test_judge_thinking_comes_from_the_settings_unless_stated(monkeypatch):from codeverse3d.judges.vlm_judge import VlmJudgefrom codeverse3d.tracks.lifecycle import BaseTrack- assert VlmJudge("scene_v2", model_id="fake:fake-1").thinking == Settings().judge.thinking == "low"+ assert VlmJudge("scene", model_id="fake:fake-1").thinking == Settings().judge.thinking == "low"s = Settings(judge={"thinking": "medium"})monkeypatch.setattr("codeverse3d.judges.vlm_judge.get_settings", lambda: s)- j = VlmJudge("scene_v2", model_id="fake:fake-1")+ j = VlmJudge("scene", model_id="fake:fake-1")assert j.thinking == "medium"- assert VlmJudge("scene_v2", model_id="fake:fake-1", thinking="high").thinking == "high" # a stated level wins+ assert VlmJudge("scene", model_id="fake:fake-1", thinking="high").thinking == "high" # a stated level winswith pytest.raises(ValueError, match="judge thinking must be one of"):- VlmJudge("scene_v2", model_id="fake:fake-1", thinking="max")+ VlmJudge("scene", model_id="fake:fake-1", thinking="max")assert BaseTrack.sampling(SimpleNamespace(track=Track.SCENE), j)["judge"]["thinking"] == "medium"get_settings.cache_clear()diff --git a/harness/tests/orchestrator_tracks/data/offrecord_runs.json b/harness/tests/orchestrator_tracks/data/offrecord_runs.jsonindex b70e0f73..4b22633a 100644--- a/harness/tests/orchestrator_tracks/data/offrecord_runs.json+++ b/harness/tests/orchestrator_tracks/data/offrecord_runs.json@@ -128,28 +128,28 @@"judges": [{"round": 0,- "rubric": "articulated_v1",+ "rubric": "articulated","cost_usd": 0.072284,"input_tokens": 11668,"output_tokens": 2335},{"round": 1,- "rubric": "articulated_v1",+ "rubric": "articulated","cost_usd": 0.071556,"input_tokens": 12102,"output_tokens": 2260},{"round": 2,- "rubric": "articulated_v1",+ "rubric": "articulated","cost_usd": 0.073526,"input_tokens": 10681,"output_tokens": 1649},{"round": 3,- "rubric": "articulated_v1",+ "rubric": "articulated","cost_usd": 0.067796,"input_tokens": 10558,"output_tokens": 1868@@ -370,14 +370,14 @@"judges": [{"round": 0,- "rubric": "static_object_v1",+ "rubric": "static_object","cost_usd": 0.044482,"input_tokens": 10802,"output_tokens": 1956},{"round": 1,- "rubric": "static_object_v1",+ "rubric": "static_object","cost_usd": 0.062174,"input_tokens": 10957,"output_tokens": 2020@@ -495,21 +495,21 @@"judges": [{"round": 0,- "rubric": "static_object_v1",+ "rubric": "static_object","cost_usd": 0.053338,"input_tokens": 10601,"output_tokens": 1639},{"round": 1,- "rubric": "static_object_v1",+ "rubric": "static_object","cost_usd": 0.062198,"input_tokens": 10705,"output_tokens": 1721},{"round": 2,- "rubric": "static_object_v1",+ "rubric": "static_object","cost_usd": 0.057404,"input_tokens": 10750,"output_tokens": 1679diff --git a/harness/tests/orchestrator_tracks/fakes.py b/harness/tests/orchestrator_tracks/fakes.pyindex 6f58b1fe..f4f54e19 100644--- a/harness/tests/orchestrator_tracks/fakes.py+++ b/harness/tests/orchestrator_tracks/fakes.py@@ -364,7 +364,7 @@ class FakeServices(Services):def judge(self, rubric: str, model_id: str, n_samples: int = 1) -> Any:return self._judge if self._judge is not None else FakeJudge()- def reference_judge(self, model_id: str, n_samples: int = 1, rubric: str = "reference_v1") -> Any:+ def reference_judge(self, model_id: str, n_samples: int = 1, rubric: str = "reference") -> Any:self.reference_judges.append((model_id, n_samples, rubric))return self._judge if self._judge is not None else FakeJudge()diff --git a/harness/tests/orchestrator_tracks/test_candidates_reference_motion.py b/harness/tests/orchestrator_tracks/test_candidates_reference_motion.pyindex 9b08fc42..415c1b61 100644--- a/harness/tests/orchestrator_tracks/test_candidates_reference_motion.py+++ b/harness/tests/orchestrator_tracks/test_candidates_reference_motion.py@@ -115,7 +115,7 @@ def test_reference_images_wire_judge_prompts_and_silhouette(tmp_path, chair_plansettings=settings, runtime=FakeRuntime(Language.BLENDER))rec = track.run(spec, ws)# reference judge + rubric- assert services.reference_judges and services.reference_judges[0][2] == "reference_v1" and rec.extra["rubric"] == "reference_v1"+ assert services.reference_judges and services.reference_judges[0][2] == "reference" and rec.extra["rubric"] == "reference"# prompts mention the references; single-shot attaches the imagefrom codeverse3d.contracts.chat import ImagePartdiff --git a/harness/tests/orchestrator_tracks/test_cost_latency.py b/harness/tests/orchestrator_tracks/test_cost_latency.pyindex 42dc5d86..4504a2f1 100644--- a/harness/tests/orchestrator_tracks/test_cost_latency.py+++ b/harness/tests/orchestrator_tracks/test_cost_latency.py@@ -142,7 +142,7 @@ def _scene_ctx(tmp_path, settings, *, services, agent=None, plan=None, agent_id=rt = FakeRuntime(Language.SCENE_THREEJS)ctx = RunContext(spec=spec, ws=ws, events=EventLog(ws.events_path), settings=settings,budget=BudgetGuard(spec.budget, soft_fraction=0.55), runtime=rt, services=services,- state=RunState(), policy=RoundPolicy(), track=Track.SCENE, rubric="scene_v2",+ state=RunState(), policy=RoundPolicy(), track=Track.SCENE, rubric="scene",agent_id=agent_id, plan=plan, agent=agent, contract_text="CONTRACT")rt.skeleton(ws, plan)ws.commit("skeleton")diff --git a/harness/tests/orchestrator_tracks/test_likeness_routing.py b/harness/tests/orchestrator_tracks/test_likeness_routing.pyindex 39d4d098..7f05c077 100644--- a/harness/tests/orchestrator_tracks/test_likeness_routing.py+++ b/harness/tests/orchestrator_tracks/test_likeness_routing.py@@ -13,7 +13,7 @@ from codeverse3d.tracks.prompting import reference_notedef _ctx(track: Track, refs: list[str], *, single_shot: bool = False) -> SimpleNamespace:spec = SimpleNamespace(references=[ReferenceImage(path=p, role="likeness", note="x") for p in refs],backends=SimpleNamespace(judge="gemini:pro"))- return SimpleNamespace(spec=spec, track=track, single_shot=single_shot, rubric="shader_v2",+ return SimpleNamespace(spec=spec, track=track, single_shot=single_shot, rubric="shader",policy=SimpleNamespace(judge_samples=2))@@ -40,17 +40,17 @@ def test_make_judge_routes_photos_to_the_likeness_judge_and_a_replay_picks_the_sassert BaseTrack.make_judge(me, _ctx(Track.SCENE, ["/x.png"])) == "L"assert BaseTrack.make_judge(me, _ctx(Track.STATIC_OBJECT, ["/x.png"])) == "R"assert BaseTrack.make_judge(me, _ctx(Track.GRAPHICS, [])) == "P"- assert calls[0] == ("likeness", "shader_v2")+ assert calls[0] == ("likeness", "shader")# a replay picks the judge the live run did: `3dcode judge` / calibration and the live run share# `judge_for` — a graphics or scene run with photos was re-judged by ReferenceJudge (silhouette +# diff pass) where the run used likenessfrom codeverse3d.cli._judge import make_judgefrom codeverse3d.judges.vlm_judge import LikenessJudge, ReferenceJudge, VlmJudge, judge_for- assert judge_for(Track.GRAPHICS, "shader_v2", references=True) is LikenessJudge- assert judge_for(Track.SCENE, "scene_v2", references=True) is LikenessJudge- assert judge_for(Track.STATIC_OBJECT, "reference_v1", references=True) is ReferenceJudge- assert judge_for(Track.STATIC_OBJECT, "reference_v1", references=False) is ReferenceJudge # measured criteria- assert judge_for(Track.GRAPHICS, "shader_v2", references=False) is VlmJudge+ assert judge_for(Track.GRAPHICS, "shader", references=True) is LikenessJudge+ assert judge_for(Track.SCENE, "scene", references=True) is LikenessJudge+ assert judge_for(Track.STATIC_OBJECT, "reference", references=True) is ReferenceJudge+ assert judge_for(Track.STATIC_OBJECT, "reference", references=False) is ReferenceJudge # measured criteria+ assert judge_for(Track.GRAPHICS, "shader", references=False) is VlmJudgespec = SimpleNamespace(track=Track.GRAPHICS, references=[ReferenceImage(path="/x.png", role="likeness", note="")])- assert type(make_judge(spec, "shader_v2", "gemini:pro", 1)) is LikenessJudge+ assert type(make_judge(spec, "shader", "gemini:pro", 1)) is LikenessJudgediff --git a/harness/tests/orchestrator_tracks/test_round_economics.py b/harness/tests/orchestrator_tracks/test_round_economics.pyindex 15d2d5d5..94e3ac76 100644--- a/harness/tests/orchestrator_tracks/test_round_economics.py+++ b/harness/tests/orchestrator_tracks/test_round_economics.py@@ -150,7 +150,7 @@ def _ctx(tmp_path, spec, settings):return RunContext(spec=spec, ws=ws, events=EventLog(ws.events_path), settings=settings,budget=BudgetGuard(Budget(max_minutes=60)), runtime=FakeRuntime(Language.THREEJS),services=FakeServices(), state=RunState(), policy=RoundPolicy(), track=spec.track,- rubric="static_object_v1", agent_id="fake:fake-model", judge=FakeJudge())+ rubric="static_object", agent_id="fake:fake-model", judge=FakeJudge())# ----------------------------------------------------------------------------- end to enddiff --git a/harness/tests/prompts/manifest.json b/harness/tests/prompts/manifest.jsonindex fa3ff094..6f19c1b2 100644--- a/harness/tests/prompts/manifest.json+++ b/harness/tests/prompts/manifest.json@@ -18,14 +18,14 @@"job:refine@r01/prompt": {"bytes": 7562, "sha256": "328c6f4ffce62283a158a78af4dd3f87b0322efdca2679da200b3b9139464e60"},"job:refine@r01/scope": {"bytes": 379, "sha256": "d17ae76997ec9443d52089e32ec81ecac3ba2c93529b227533942e84598113d9"},"job:refine@r01/system": {"bytes": 3271, "sha256": "29f5a9b2e7b9a879be58963ce95ae6fbec7460dae6785d542b214d6b850e21f4"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9884, "sha256": "b8b8e6c5f11f440a8563eb9bbb1c2dd9fd32f78da6d35078e1f54eb231cc80c0"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 7266, "sha256": "44869437393997f2107fdf0a59b70d830557064fa5958abfbee786e535323a2c"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/system": {"bytes": 13132, "sha256": "559ab1621aef1f28ee9c9d235bf78a1003cb94358ed85248cf0d652bb40498d1"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10787, "sha256": "ccb99967b785a412cbd6bdbf0b1ecfc92cb5ba25747e678bbcc6a8c4e56b83e1"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7591, "sha256": "a2e7394b1d8515b0256e914ceb6b07eaf526a634104ea03829c559e5fd7ebf67"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/system": {"bytes": 13132, "sha256": "559ab1621aef1f28ee9c9d235bf78a1003cb94358ed85248cf0d652bb40498d1"},+ "judge:articulated:runs/run/artifacts/renders/r00/input": {"bytes": 9884, "sha256": "b8b8e6c5f11f440a8563eb9bbb1c2dd9fd32f78da6d35078e1f54eb231cc80c0"},+ "judge:articulated:runs/run/artifacts/renders/r00/messages": {"bytes": 7266, "sha256": "44869437393997f2107fdf0a59b70d830557064fa5958abfbee786e535323a2c"},+ "judge:articulated:runs/run/artifacts/renders/r00/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},+ "judge:articulated:runs/run/artifacts/renders/r00/system": {"bytes": 13129, "sha256": "b6c7c917b035224ca53b1a3e1fcd0c838b8ac20563e687cd6521d8cab4a78ed6"},+ "judge:articulated:runs/run/artifacts/renders/r01/input": {"bytes": 10787, "sha256": "ccb99967b785a412cbd6bdbf0b1ecfc92cb5ba25747e678bbcc6a8c4e56b83e1"},+ "judge:articulated:runs/run/artifacts/renders/r01/messages": {"bytes": 7591, "sha256": "a2e7394b1d8515b0256e914ceb6b07eaf526a634104ea03829c559e5fd7ebf67"},+ "judge:articulated:runs/run/artifacts/renders/r01/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},+ "judge:articulated:runs/run/artifacts/renders/r01/system": {"bytes": 13129, "sha256": "b6c7c917b035224ca53b1a3e1fcd0c838b8ac20563e687cd6521d8cab4a78ed6"},"skills:runs/run:baseline@r00/routed": {"bytes": 125, "sha256": "7e099a4ebdaae69160ddbaf1be8d8c7349914d676244b57b1815508b507fd177"},"skills:runs/run:refine@r01/routed": {"bytes": 125, "sha256": "5c725d8f66ce509d26245268d4c3525c99ce65ff9a56b94c708a74934ae2edf4"}},@@ -51,14 +51,14 @@"chat:planner:planner/messages": {"bytes": 1328, "sha256": "313c87553cf38df423dc9245a8af4cca5a41a844b489e13a5dc08e8c44eabba5"},"chat:planner:planner/sampling": {"bytes": 10001, "sha256": "0ff37a3ae43fdf878dd0911a9954d57e49365ad788c84818ea01d350310ab401"},"chat:planner:planner/system": {"bytes": 20030, "sha256": "45dccfcf72b4c6e46b7ceeaae8af0ebddf6f6e6b5970cda5fe900aa47682bf06"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9604, "sha256": "80435bd0207e04f9d3384039b4b2c140056e8406af4c3616cfc019430969244b"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 7120, "sha256": "7a0ef8d51456faa2c38867eea6c502fbe1a7aa30d6cd9f23b44701a2d11846bf"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},- "judge:articulated_v1:runs/run/artifacts/renders/r00/system": {"bytes": 13132, "sha256": "559ab1621aef1f28ee9c9d235bf78a1003cb94358ed85248cf0d652bb40498d1"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10507, "sha256": "1dc1b1c082595444ef377d6bce19d98357d14450f414fca3dc0e80728a12da8d"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7445, "sha256": "2a654e5bb10f8a642367b40c8893a16dce24c4bc4f6941cfbba76ce348ad4f1c"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},- "judge:articulated_v1:runs/run/artifacts/renders/r01/system": {"bytes": 13132, "sha256": "559ab1621aef1f28ee9c9d235bf78a1003cb94358ed85248cf0d652bb40498d1"},+ "judge:articulated:runs/run/artifacts/renders/r00/input": {"bytes": 9604, "sha256": "80435bd0207e04f9d3384039b4b2c140056e8406af4c3616cfc019430969244b"},+ "judge:articulated:runs/run/artifacts/renders/r00/messages": {"bytes": 7120, "sha256": "7a0ef8d51456faa2c38867eea6c502fbe1a7aa30d6cd9f23b44701a2d11846bf"},+ "judge:articulated:runs/run/artifacts/renders/r00/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},+ "judge:articulated:runs/run/artifacts/renders/r00/system": {"bytes": 13129, "sha256": "b6c7c917b035224ca53b1a3e1fcd0c838b8ac20563e687cd6521d8cab4a78ed6"},+ "judge:articulated:runs/run/artifacts/renders/r01/input": {"bytes": 10507, "sha256": "1dc1b1c082595444ef377d6bce19d98357d14450f414fca3dc0e80728a12da8d"},+ "judge:articulated:runs/run/artifacts/renders/r01/messages": {"bytes": 7445, "sha256": "2a654e5bb10f8a642367b40c8893a16dce24c4bc4f6941cfbba76ce348ad4f1c"},+ "judge:articulated:runs/run/artifacts/renders/r01/sampling": {"bytes": 11989, "sha256": "e93c46f5a7a740cd7cf30206976488736804f2b016f39424b2f0e41bfbea1c29"},+ "judge:articulated:runs/run/artifacts/renders/r01/system": {"bytes": 13129, "sha256": "b6c7c917b035224ca53b1a3e1fcd0c838b8ac20563e687cd6521d8cab4a78ed6"},"skills:runs/run:baseline@r00/routed": {"bytes": 138, "sha256": "cb0610ba263bbb5cbeadf679a8213b52ece498906fabcdf2f8e5937d23c94c0b"},"skills:runs/run:refine@r01/routed": {"bytes": 138, "sha256": "cb0610ba263bbb5cbeadf679a8213b52ece498906fabcdf2f8e5937d23c94c0b"}},@@ -71,28 +71,28 @@"payloads": {"chat:planner:planner/messages": {"bytes": 348, "sha256": "f50fc3cdf3a426fc9c9b830b6e68d2cfe9fd9b7125550362b018f9c9b454b25c"},"chat:planner:planner/sampling": {"bytes": 3421, "sha256": "918170272cea656e8c542c54b777c57d1a4567a8cc36a0c1cd825cb81a01b9bb"},- "chat:planner:planner/system": {"bytes": 11922, "sha256": "172a4deaa1499e7ee80be931e5840b293ed81ee0a74262e20459261d81d40f43"},- "context:runs/run/_cand/c0/files": {"bytes": 32162, "sha256": "37bbde6b0aeceba63e2706842f0598cbeed13e67efa8dbd1f5c75f08248e9ce6"},- "context:runs/run/_cand/c1/files": {"bytes": 32162, "sha256": "37bbde6b0aeceba63e2706842f0598cbeed13e67efa8dbd1f5c75f08248e9ce6"},- "context:runs/run/files": {"bytes": 32162, "sha256": "37bbde6b0aeceba63e2706842f0598cbeed13e67efa8dbd1f5c75f08248e9ce6"},- "job:baseline_c0@r00/prompt": {"bytes": 11359, "sha256": "d55a46da0bbbd9106773fd9d0664b2f52ac44d84988a5d3dac1bc5e66cdea589"},+ "chat:planner:planner/system": {"bytes": 11919, "sha256": "194d0e2a724c4799816033228af15cac568dd51c14ed85bff9c94ae5ba281866"},+ "context:runs/run/_cand/c0/files": {"bytes": 32156, "sha256": "072c694a9a953e9ac63302ccd0b3165f4560e3f104d16cdda58b65bb840f82c0"},+ "context:runs/run/_cand/c1/files": {"bytes": 32156, "sha256": "072c694a9a953e9ac63302ccd0b3165f4560e3f104d16cdda58b65bb840f82c0"},+ "context:runs/run/files": {"bytes": 32156, "sha256": "072c694a9a953e9ac63302ccd0b3165f4560e3f104d16cdda58b65bb840f82c0"},+ "job:baseline_c0@r00/prompt": {"bytes": 11356, "sha256": "5e82c0438d0d084acbf0d2cfacfdde06ef2c7b9cd5675e258ac5a2c032a82fee"},"job:baseline_c0@r00/scope": {"bytes": 383, "sha256": "51d22961200ef987a7c652af48ff2897b910803baa672d3081f53c3fb819a685"},- "job:baseline_c0@r00/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},- "job:baseline_c1@r00/prompt": {"bytes": 11359, "sha256": "d55a46da0bbbd9106773fd9d0664b2f52ac44d84988a5d3dac1bc5e66cdea589"},+ "job:baseline_c0@r00/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},+ "job:baseline_c1@r00/prompt": {"bytes": 11356, "sha256": "5e82c0438d0d084acbf0d2cfacfdde06ef2c7b9cd5675e258ac5a2c032a82fee"},"job:baseline_c1@r00/scope": {"bytes": 383, "sha256": "831265b54627f1f6c1bfadebefa4e898d6b543b253bf6be0685b5383bd39fb57"},- "job:baseline_c1@r00/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},- "judge:shader_v2:runs/run/_cand/c0/artifacts/renders/r00/input": {"bytes": 4078, "sha256": "e3dd787e6b7cc0aab8d35ce01ff5fbd42a3da389411b7402e6b15a0301f21f8a"},- "judge:shader_v2:runs/run/_cand/c0/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/_cand/c0/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/_cand/c0/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/_cand/c1/artifacts/renders/r00/input": {"bytes": 4078, "sha256": "c825f0fbe74ca0c4fd90b8b3ff2bc2177851d4598ef3878e6ed7753d87d772a3"},- "judge:shader_v2:runs/run/_cand/c1/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/_cand/c1/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/_cand/c1/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/artifacts/renders/r00/input": {"bytes": 4024, "sha256": "f7a7cf02a2ac234aa62fc4ab4e3be6cb6cf2305a24c07c66bafec8e728038efd"},- "judge:shader_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},+ "job:baseline_c1@r00/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},+ "judge:shader:runs/run/_cand/c0/artifacts/renders/r00/input": {"bytes": 4078, "sha256": "e3dd787e6b7cc0aab8d35ce01ff5fbd42a3da389411b7402e6b15a0301f21f8a"},+ "judge:shader:runs/run/_cand/c0/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/_cand/c0/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/_cand/c0/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/_cand/c1/artifacts/renders/r00/input": {"bytes": 4078, "sha256": "c825f0fbe74ca0c4fd90b8b3ff2bc2177851d4598ef3878e6ed7753d87d772a3"},+ "judge:shader:runs/run/_cand/c1/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/_cand/c1/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/_cand/c1/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/artifacts/renders/r00/input": {"bytes": 4024, "sha256": "f7a7cf02a2ac234aa62fc4ab4e3be6cb6cf2305a24c07c66bafec8e728038efd"},+ "judge:shader:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},"skills:runs/run/_cand/c0:baseline@r00/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"},"skills:runs/run/_cand/c1:baseline@r00/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"},"skills:runs/run:baseline@r00/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"}@@ -106,29 +106,29 @@"payloads": {"chat:planner:planner/messages": {"bytes": 348, "sha256": "f50fc3cdf3a426fc9c9b830b6e68d2cfe9fd9b7125550362b018f9c9b454b25c"},"chat:planner:planner/sampling": {"bytes": 3421, "sha256": "918170272cea656e8c542c54b777c57d1a4567a8cc36a0c1cd825cb81a01b9bb"},- "chat:planner:planner/system": {"bytes": 11922, "sha256": "172a4deaa1499e7ee80be931e5840b293ed81ee0a74262e20459261d81d40f43"},- "context:runs/run/files": {"bytes": 32162, "sha256": "37bbde6b0aeceba63e2706842f0598cbeed13e67efa8dbd1f5c75f08248e9ce6"},- "job:baseline@r00/prompt": {"bytes": 11359, "sha256": "d55a46da0bbbd9106773fd9d0664b2f52ac44d84988a5d3dac1bc5e66cdea589"},+ "chat:planner:planner/system": {"bytes": 11919, "sha256": "194d0e2a724c4799816033228af15cac568dd51c14ed85bff9c94ae5ba281866"},+ "context:runs/run/files": {"bytes": 32156, "sha256": "072c694a9a953e9ac63302ccd0b3165f4560e3f104d16cdda58b65bb840f82c0"},+ "job:baseline@r00/prompt": {"bytes": 11356, "sha256": "5e82c0438d0d084acbf0d2cfacfdde06ef2c7b9cd5675e258ac5a2c032a82fee"},"job:baseline@r00/scope": {"bytes": 379, "sha256": "831e9c99bdd04789b89bfddb89c97939a813323c9173d742adabbb5f1f8a7dc1"},- "job:baseline@r00/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},+ "job:baseline@r00/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},"job:refine@r01/prompt": {"bytes": 4829, "sha256": "701203307b2677078ab21caca6c39556b594c220e1642a242e45fad4caf905f1"},"job:refine@r01/scope": {"bytes": 375, "sha256": "f106c6fcaee6849acebc6005bde1bad2939260265ed9c794846543d16e84edce"},- "job:refine@r01/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},+ "job:refine@r01/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},"job:refine@r02/prompt": {"bytes": 4828, "sha256": "f02f3a218edb95d79000fd339cc556045baccd5087231dd293db83195c927fa7"},"job:refine@r02/scope": {"bytes": 375, "sha256": "c4a7fdb0b390ec75d595006e11b9982113e60c49f5a62a978153167053bbcd33"},- "job:refine@r02/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},- "judge:shader_v2:runs/run/artifacts/renders/r00/input": {"bytes": 4024, "sha256": "77a8a69c16f42514152b7169c1afc91bca46d4ececc6bf60e826020cfb87af09"},- "judge:shader_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/artifacts/renders/r01/input": {"bytes": 5077, "sha256": "91baae5de582cf7919bad97282b84d07db84fed38e5a2304e901dfd2a2748b0a"},- "judge:shader_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 4123, "sha256": "a1f0c207406de6f415fdb8516b695aaedf7f06d25598f33932d2ad728f8c881f"},- "judge:shader_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r01/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/artifacts/renders/r02/input": {"bytes": 5073, "sha256": "f441039d9cbbfd50ed5d1a38fb008fc63e2c19ed191a866351532a6a190e1c35"},- "judge:shader_v2:runs/run/artifacts/renders/r02/messages": {"bytes": 4123, "sha256": "5b81b09f9d6a7e3225549454099d07143f5bb726c3ccff263b4afc7bfe13d88b"},- "judge:shader_v2:runs/run/artifacts/renders/r02/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r02/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},+ "job:refine@r02/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},+ "judge:shader:runs/run/artifacts/renders/r00/input": {"bytes": 4024, "sha256": "77a8a69c16f42514152b7169c1afc91bca46d4ececc6bf60e826020cfb87af09"},+ "judge:shader:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/artifacts/renders/r01/input": {"bytes": 5077, "sha256": "91baae5de582cf7919bad97282b84d07db84fed38e5a2304e901dfd2a2748b0a"},+ "judge:shader:runs/run/artifacts/renders/r01/messages": {"bytes": 4123, "sha256": "a1f0c207406de6f415fdb8516b695aaedf7f06d25598f33932d2ad728f8c881f"},+ "judge:shader:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r01/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/artifacts/renders/r02/input": {"bytes": 5073, "sha256": "f441039d9cbbfd50ed5d1a38fb008fc63e2c19ed191a866351532a6a190e1c35"},+ "judge:shader:runs/run/artifacts/renders/r02/messages": {"bytes": 4123, "sha256": "5b81b09f9d6a7e3225549454099d07143f5bb726c3ccff263b4afc7bfe13d88b"},+ "judge:shader:runs/run/artifacts/renders/r02/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r02/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},"skills:runs/run:baseline@r00/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"},"skills:runs/run:refine@r01/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"},"skills:runs/run:refine@r02/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"}@@ -142,24 +142,24 @@"payloads": {"chat:planner:planner/messages": {"bytes": 348, "sha256": "f50fc3cdf3a426fc9c9b830b6e68d2cfe9fd9b7125550362b018f9c9b454b25c"},"chat:planner:planner/sampling": {"bytes": 3421, "sha256": "918170272cea656e8c542c54b777c57d1a4567a8cc36a0c1cd825cb81a01b9bb"},- "chat:planner:planner/system": {"bytes": 11922, "sha256": "172a4deaa1499e7ee80be931e5840b293ed81ee0a74262e20459261d81d40f43"},- "context:runs/run/files": {"bytes": 32162, "sha256": "37bbde6b0aeceba63e2706842f0598cbeed13e67efa8dbd1f5c75f08248e9ce6"},- "job:baseline@r00/prompt": {"bytes": 11359, "sha256": "d55a46da0bbbd9106773fd9d0664b2f52ac44d84988a5d3dac1bc5e66cdea589"},+ "chat:planner:planner/system": {"bytes": 11919, "sha256": "194d0e2a724c4799816033228af15cac568dd51c14ed85bff9c94ae5ba281866"},+ "context:runs/run/files": {"bytes": 32156, "sha256": "072c694a9a953e9ac63302ccd0b3165f4560e3f104d16cdda58b65bb840f82c0"},+ "job:baseline@r00/prompt": {"bytes": 11356, "sha256": "5e82c0438d0d084acbf0d2cfacfdde06ef2c7b9cd5675e258ac5a2c032a82fee"},"job:baseline@r00/scope": {"bytes": 379, "sha256": "831e9c99bdd04789b89bfddb89c97939a813323c9173d742adabbb5f1f8a7dc1"},- "job:baseline@r00/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},+ "job:baseline@r00/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},"job:r00_baseline_repair1@r00/prompt": {"bytes": 2407, "sha256": "056a91d52ba3c3b09efd2fe13f139c6953fe99205e7c7b53ffc26331b1701b72"},"job:r00_baseline_repair1@r00/scope": {"bytes": 389, "sha256": "6e9ef696a5ada34a890d14e7cd2de9c69d51e95e9a36bd0d8d9db7b043e06671"},- "job:r00_baseline_repair1@r00/system": {"bytes": 9640, "sha256": "253de140b4d3eba01e244c0dd785548f1241a91dc42495e34a98cdcd3284b4fc"},+ "job:r00_baseline_repair1@r00/system": {"bytes": 9634, "sha256": "53029a91abfcd13198024a81e77e6b50003f92a850c6c0e339e268a1a22319f9"},"job:r00_baseline_repair2@r00/prompt": {"bytes": 2712, "sha256": "aa2cfd9457a008820bb834e4931277ad4e7d8826de45de18fd13c21045617991"},"job:r00_baseline_repair2@r00/scope": {"bytes": 391, "sha256": "7e0dbd43421db95ef9078ca03fd146cccc2cc8372380a72fd55fa12ff7217db0"},- "job:r00_baseline_repair2@r00/system": {"bytes": 9640, "sha256": "253de140b4d3eba01e244c0dd785548f1241a91dc42495e34a98cdcd3284b4fc"},- "job:rebuild@r01/prompt": {"bytes": 11542, "sha256": "79d440c92dc02752e64ab8d61dde2ae6bd75c737e26d9ed0f9da75e12f6080ff"},+ "job:r00_baseline_repair2@r00/system": {"bytes": 9634, "sha256": "53029a91abfcd13198024a81e77e6b50003f92a850c6c0e339e268a1a22319f9"},+ "job:rebuild@r01/prompt": {"bytes": 11539, "sha256": "1feb154736db6ae50413fddd50641bdac992613a50bdc769980d515c842fd433"},"job:rebuild@r01/scope": {"bytes": 379, "sha256": "471fba995682e47fe12099cb5881743a8aeae71a0243fa3598c72d69a5b31f98"},- "job:rebuild@r01/system": {"bytes": 3176, "sha256": "3483d6a17a75ef02d53f4addcc8713e0b7e281c092fa88a854524c2bae344f32"},- "judge:shader_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4024, "sha256": "f6637594bb85a4806d5e768a6467825652abb7855d33ae866733714bf4489f62"},- "judge:shader_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r01/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},+ "job:rebuild@r01/system": {"bytes": 3173, "sha256": "c3879a563e2a512b82358d0c3927dcbfe9347725e592035a9b64c5d1b34e6654"},+ "judge:shader:runs/run/artifacts/renders/r01/input": {"bytes": 4024, "sha256": "f6637594bb85a4806d5e768a6467825652abb7855d33ae866733714bf4489f62"},+ "judge:shader:runs/run/artifacts/renders/r01/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r01/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},"skills:runs/run:baseline@r00/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"},"skills:runs/run:refine@r01/routed": {"bytes": 56, "sha256": "400eb18b3e0ca869392d11f9fc95c12140d8234f3efe8f3c1ec66b5c43d1ffa6"}},@@ -170,23 +170,23 @@},"graphics.glsl_shader.single_shot.fresh": {"payloads": {- "chat:model:baseline:r00/messages": {"bytes": 20198, "sha256": "cd7e4a7b6be62c421c035311b610f14624ab9c66f2f30d475004c136273565f8"},+ "chat:model:baseline:r00/messages": {"bytes": 20189, "sha256": "ad6add1a38608c8e4031e76136924909bac84377f9b9e820f0244a48a0b49577"},"chat:model:baseline:r00/sampling": {"bytes": 92, "sha256": "ed52acf6f9a2dcc4c97fd8698b845dfed301b15f5fc05262bd934a525b14fc9b"},- "chat:model:baseline:r00/system": {"bytes": 5518, "sha256": "4692c1cfe4fedae824d44f13ea4902d1139b025c5c4d565c26332a6fd2f8f0f3"},- "chat:model:refine:r01/messages": {"bytes": 11028, "sha256": "9b8c9a691c8d022fe9052ed4cc9a8dbcc87bf7887b239d97a2703cedcb069c4c"},+ "chat:model:baseline:r00/system": {"bytes": 5515, "sha256": "ab316b530b24a3c7bdcd5076264e7b392e88e7b4af4f78708381c12d137895d1"},+ "chat:model:refine:r01/messages": {"bytes": 11022, "sha256": "554d156198784aaada06eeaf4bef6233978ff983cb634fab6edb52a8903f273b"},"chat:model:refine:r01/sampling": {"bytes": 92, "sha256": "08e72af1b4fb249378c5a151e019ded70a1014eafa4d50f70927bb225f695678"},- "chat:model:refine:r01/system": {"bytes": 5518, "sha256": "4692c1cfe4fedae824d44f13ea4902d1139b025c5c4d565c26332a6fd2f8f0f3"},+ "chat:model:refine:r01/system": {"bytes": 5515, "sha256": "ab316b530b24a3c7bdcd5076264e7b392e88e7b4af4f78708381c12d137895d1"},"chat:planner:planner/messages": {"bytes": 348, "sha256": "f50fc3cdf3a426fc9c9b830b6e68d2cfe9fd9b7125550362b018f9c9b454b25c"},"chat:planner:planner/sampling": {"bytes": 3421, "sha256": "918170272cea656e8c542c54b777c57d1a4567a8cc36a0c1cd825cb81a01b9bb"},- "chat:planner:planner/system": {"bytes": 11922, "sha256": "172a4deaa1499e7ee80be931e5840b293ed81ee0a74262e20459261d81d40f43"},- "judge:shader_v2:runs/run/artifacts/renders/r00/input": {"bytes": 4041, "sha256": "b1c2c219851e29fdf3e2ef55acdcbd05f50f1306cb15e643990d5c550a860a0f"},- "judge:shader_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},- "judge:shader_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/artifacts/renders/r01/input": {"bytes": 5094, "sha256": "329efb6f51153caef8363b23519c03cff958df73f76dd6a6f0eb4a539efb97f2"},- "judge:shader_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 4123, "sha256": "a1f0c207406de6f415fdb8516b695aaedf7f06d25598f33932d2ad728f8c881f"},- "judge:shader_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r01/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},+ "chat:planner:planner/system": {"bytes": 11919, "sha256": "194d0e2a724c4799816033228af15cac568dd51c14ed85bff9c94ae5ba281866"},+ "judge:shader:runs/run/artifacts/renders/r00/input": {"bytes": 4041, "sha256": "b1c2c219851e29fdf3e2ef55acdcbd05f50f1306cb15e643990d5c550a860a0f"},+ "judge:shader:runs/run/artifacts/renders/r00/messages": {"bytes": 3727, "sha256": "d7deb2c4a1fa8e9b015a4e81439f9781e7618be6a01b013d18bb3acc4f6a32b9"},+ "judge:shader:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/artifacts/renders/r01/input": {"bytes": 5094, "sha256": "329efb6f51153caef8363b23519c03cff958df73f76dd6a6f0eb4a539efb97f2"},+ "judge:shader:runs/run/artifacts/renders/r01/messages": {"bytes": 4123, "sha256": "a1f0c207406de6f415fdb8516b695aaedf7f06d25598f33932d2ad728f8c881f"},+ "judge:shader:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r01/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},"skills:runs/run:baseline@r00/routed": {"bytes": 68, "sha256": "cf6a7b5027626c144cf9f6ccc254f9bf6332217398c364638db88002139012ad"},"skills:runs/run:refine@r01/routed": {"bytes": 68, "sha256": "cf6a7b5027626c144cf9f6ccc254f9bf6332217398c364638db88002139012ad"}},@@ -207,14 +207,14 @@"job:refine@r01/prompt": {"bytes": 4809, "sha256": "564c35401c05213d9906d0ecf84ead9e8c8a54ef51cae13eeb59fcb98e1b7275"},"job:refine@r01/scope": {"bytes": 354, "sha256": "81a4042ba077832530b2a9f64d3e1415da1b655b00caf1006ed51ee616a7b17b"},"job:refine@r01/system": {"bytes": 2729, "sha256": "d12d4891ce72156aa561bdfa0ed1fb20c5826412024b0be07b423949528e5df6"},- "judge:shader_v2:runs/run/artifacts/renders/r00/input": {"bytes": 4028, "sha256": "c73efe7b56d4885794e8bc52178cd5d56c0c1edc71cae0eb72159619e69e6d75"},- "judge:shader_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 3731, "sha256": "355b5a5eb7617171897155fe818a3975dea97d127ae8416f1352a67bb6a84a19"},- "judge:shader_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r00/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},- "judge:shader_v2:runs/run/artifacts/renders/r01/input": {"bytes": 5081, "sha256": "d780a48e47e0701679920bd7b03fb703275968f4d36ee59e257453d45942a00c"},- "judge:shader_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 4127, "sha256": "a4c4a41eea8d4af3df6646d2aa757c01a0ab663614361e61e76e84e21f9d5378"},- "judge:shader_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},- "judge:shader_v2:runs/run/artifacts/renders/r01/system": {"bytes": 11256, "sha256": "27c834df165bcb2e2406b0045b141fd290cf85edc1fc104a1e95b959d51aec57"},+ "judge:shader:runs/run/artifacts/renders/r00/input": {"bytes": 4028, "sha256": "c73efe7b56d4885794e8bc52178cd5d56c0c1edc71cae0eb72159619e69e6d75"},+ "judge:shader:runs/run/artifacts/renders/r00/messages": {"bytes": 3731, "sha256": "355b5a5eb7617171897155fe818a3975dea97d127ae8416f1352a67bb6a84a19"},+ "judge:shader:runs/run/artifacts/renders/r00/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r00/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},+ "judge:shader:runs/run/artifacts/renders/r01/input": {"bytes": 5081, "sha256": "d780a48e47e0701679920bd7b03fb703275968f4d36ee59e257453d45942a00c"},+ "judge:shader:runs/run/artifacts/renders/r01/messages": {"bytes": 4127, "sha256": "a4c4a41eea8d4af3df6646d2aa757c01a0ab663614361e61e76e84e21f9d5378"},+ "judge:shader:runs/run/artifacts/renders/r01/sampling": {"bytes": 11659, "sha256": "2f41a011f9775b7aaf1ec8c1594be4e9172d23f6eb6dfc4b262c6e6bf1097792"},+ "judge:shader:runs/run/artifacts/renders/r01/system": {"bytes": 11253, "sha256": "dcf75b56942831a1db5074023dfdad4b4750a3842eb4233dc2d41adacd1de67b"},"skills:runs/run:baseline@r00/routed": {"bytes": 61, "sha256": "6cf132dce307379276e442875a061e8e0d099d50b70a8d18d4622abd9ce8ee8c"},"skills:runs/run:refine@r01/routed": {"bytes": 61, "sha256": "6cf132dce307379276e442875a061e8e0d099d50b70a8d18d4622abd9ce8ee8c"}},@@ -281,18 +281,18 @@"job:zones_quay_water@r00/prompt": {"bytes": 50923, "sha256": "0f6b6c68040e7d87ba4125f9dcd91823cfee0e47d6849aa41a3f1c4e0d3f4c25"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "d7cde9503b57c30d54062b7f12f1023f1510fad249cff287c3b771c306a065bf"},"job:zones_quay_water@r00/system": {"bytes": 6011, "sha256": "84a58b966fee66290cfdf85997e8ae0f25f840e457f4b331cac118873d2c5cef"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "11c9b8092135f98b61dfe0e47b77561a374ab89aaa22ed9472b9b731e3bf8e7f"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "3788f8a25412e088c56476fd218179fa5dd0704ab1b8da7ae9dc8eb5551647e1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "35679341c294b774ac1c228d9a2bc56f4792f8831dd1c216b513e4427e6cb0c1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r01/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r02/input": {"bytes": 4751, "sha256": "0afb2dd38366bbb0bbb539c943ef4eeeaf4bea82f8748de687eda5af887da0fe"},- "judge:scene_v2:runs/run/artifacts/renders/r02/messages": {"bytes": 5298, "sha256": "0b15d27d13b64c082c86c2e4cea5d85a7f0b3e9daef93b5cb62f698d06570437"},- "judge:scene_v2:runs/run/artifacts/renders/r02/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r02/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "11c9b8092135f98b61dfe0e47b77561a374ab89aaa22ed9472b9b731e3bf8e7f"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "3788f8a25412e088c56476fd218179fa5dd0704ab1b8da7ae9dc8eb5551647e1"},+ "judge:scene:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "35679341c294b774ac1c228d9a2bc56f4792f8831dd1c216b513e4427e6cb0c1"},+ "judge:scene:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r01/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r02/input": {"bytes": 4751, "sha256": "0afb2dd38366bbb0bbb539c943ef4eeeaf4bea82f8748de687eda5af887da0fe"},+ "judge:scene:runs/run/artifacts/renders/r02/messages": {"bytes": 5298, "sha256": "0b15d27d13b64c082c86c2e4cea5d85a7f0b3e9daef93b5cb62f698d06570437"},+ "judge:scene:runs/run/artifacts/renders/r02/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r02/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 35, "sha256": "607ad2fd09f3e6a0617f5d52cff3edf2a91bd286e8bf5b18d5f7c0e5f6dfff80"},"skills:runs/run:env@r00/routed": {"bytes": 122, "sha256": "b71c5d72d6d4265e0a320839c1e742785860c6103adf2bf6b62f85c6c7739d0c"},"skills:runs/run:refine@r01/routed": {"bytes": 122, "sha256": "b71c5d72d6d4265e0a320839c1e742785860c6103adf2bf6b62f85c6c7739d0c"},@@ -342,14 +342,14 @@"job:zones_quay_water@r00/prompt": {"bytes": 38696, "sha256": "8524b350a4485277de24641c0a67881e56b7c25879100cee0210417af6403a62"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "d7cde9503b57c30d54062b7f12f1023f1510fad249cff287c3b771c306a065bf"},"job:zones_quay_water@r00/system": {"bytes": 7284, "sha256": "13fb6151a371cd11e8892f5b046bca581b917ccc1a267fef0aa4c189b26983f3"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "11c9b8092135f98b61dfe0e47b77561a374ab89aaa22ed9472b9b731e3bf8e7f"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "3788f8a25412e088c56476fd218179fa5dd0704ab1b8da7ae9dc8eb5551647e1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "35679341c294b774ac1c228d9a2bc56f4792f8831dd1c216b513e4427e6cb0c1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r01/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "11c9b8092135f98b61dfe0e47b77561a374ab89aaa22ed9472b9b731e3bf8e7f"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "3788f8a25412e088c56476fd218179fa5dd0704ab1b8da7ae9dc8eb5551647e1"},+ "judge:scene:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "35679341c294b774ac1c228d9a2bc56f4792f8831dd1c216b513e4427e6cb0c1"},+ "judge:scene:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r01/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 35, "sha256": "607ad2fd09f3e6a0617f5d52cff3edf2a91bd286e8bf5b18d5f7c0e5f6dfff80"},"skills:runs/run:env@r00/routed": {"bytes": 122, "sha256": "b71c5d72d6d4265e0a320839c1e742785860c6103adf2bf6b62f85c6c7739d0c"},"skills:runs/run:refine@r01/routed": {"bytes": 122, "sha256": "b71c5d72d6d4265e0a320839c1e742785860c6103adf2bf6b62f85c6c7739d0c"},@@ -403,10 +403,10 @@"chat:planner:zone-layout#2/messages": {"bytes": 485, "sha256": "67e917190ee227c5ed3415ab4b03c072af313d7040704e7de671e63c42e2ab52"},"chat:planner:zone-layout#2/sampling": {"bytes": 3397, "sha256": "160445a83ede28afc0191547fd42d9341e694ef70aaad380bebf60ead0a65c49"},"chat:planner:zone-layout#2/system": {"bytes": 3326, "sha256": "494acec302dd23fa2fc6ee1eb6d48e873d6cf0ad98bad413e27578cf062313d4"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3772, "sha256": "ff1c9893cb3e5f4d8f726e4d35fd23a0532523af32df0ca593728608b89a34c3"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3772, "sha256": "ff1c9893cb3e5f4d8f726e4d35fd23a0532523af32df0ca593728608b89a34c3"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "676859b5172ea33c86195b543b5c80e49dc907598dbdddb57a9734e58df9e8a1"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 35, "sha256": "607ad2fd09f3e6a0617f5d52cff3edf2a91bd286e8bf5b18d5f7c0e5f6dfff80"},"skills:runs/run:env@r00/routed": {"bytes": 143, "sha256": "a2ca52eb478b4f896bec65a65d4385cc8c79db85d5f827f556ded75fa5812d5f"},"skills:runs/run:refine@r01/routed": {"bytes": 143, "sha256": "a2ca52eb478b4f896bec65a65d4385cc8c79db85d5f827f556ded75fa5812d5f"},@@ -467,18 +467,18 @@"job:zones_quay_water@r00/prompt": {"bytes": 53496, "sha256": "82b80505d5bb0ecc0ed5c7bc53f7d0bafdb8dc8102cb51c314e6628b3e7ffbcc"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "5b1def419a1370ab2d9f2345bc4a5232ef3c793d4e4f1e01bd5ab9a3a8e88a7d"},"job:zones_quay_water@r00/system": {"bytes": 5614, "sha256": "b23b5fa6365b226f94f9cf4e9f3f7235628d22c1bee3ca22c6f44406ce518cdb"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "7fa5ae748295e277f0edb9fc51a241ac20ccf6c64a88431cb95bf08313fb661a"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},- "judge:scene_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r01/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r02/input": {"bytes": 4751, "sha256": "2300148c08e41d46deca89c02bad3af6d0f0170538b7ab50c270c8ad09977938"},- "judge:scene_v2:runs/run/artifacts/renders/r02/messages": {"bytes": 5298, "sha256": "d8e7af12d299bcc4d09f7b68a6cdf3de2816e89d0d785ff88153c9e3aa5ce8cd"},- "judge:scene_v2:runs/run/artifacts/renders/r02/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r02/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "7fa5ae748295e277f0edb9fc51a241ac20ccf6c64a88431cb95bf08313fb661a"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},+ "judge:scene:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},+ "judge:scene:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r01/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r02/input": {"bytes": 4751, "sha256": "2300148c08e41d46deca89c02bad3af6d0f0170538b7ab50c270c8ad09977938"},+ "judge:scene:runs/run/artifacts/renders/r02/messages": {"bytes": 5298, "sha256": "d8e7af12d299bcc4d09f7b68a6cdf3de2816e89d0d785ff88153c9e3aa5ce8cd"},+ "judge:scene:runs/run/artifacts/renders/r02/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r02/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},"skills:runs/run:env@r00/routed": {"bytes": 164, "sha256": "eb0e2aca5a76b26e49a2d2c862a9405ef2e3e1130e719cca9a1f23fe1b264402"},"skills:runs/run:refine@r01/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},@@ -541,18 +541,18 @@"job:zones_quay_water@r00/prompt": {"bytes": 53676, "sha256": "3b2b4db061b894aca9121f762690c234369d83ea0d960bb5172ad2f4a5684726"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "5b1def419a1370ab2d9f2345bc4a5232ef3c793d4e4f1e01bd5ab9a3a8e88a7d"},"job:zones_quay_water@r00/system": {"bytes": 5614, "sha256": "b23b5fa6365b226f94f9cf4e9f3f7235628d22c1bee3ca22c6f44406ce518cdb"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/input": {"bytes": 3824, "sha256": "d2b8392354ef8ee0eaaa261eedb39f1c73fddc4373960c48d513e50b707daa57"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/messages": {"bytes": 3099, "sha256": "eee511c765a105054417cb3fa37bf67c24bcf9022da44dc086385878a61d5f67"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/sampling": {"bytes": 8355, "sha256": "a80299c8dfc30739d05a0274526c361cb406c78a1ab5e923fcdf78d8c78ea73a"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/system": {"bytes": 8837, "sha256": "6dbacdd03ad0deab9f9264199b854eb37d8c29b58b5eaab8d1e78544f319a4f5"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/input": {"bytes": 3844, "sha256": "f53c37a588a498c52a403ce9178c0aea6e599a0b8ec944c1a9a88baf0e6ef175"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/messages": {"bytes": 3099, "sha256": "eee511c765a105054417cb3fa37bf67c24bcf9022da44dc086385878a61d5f67"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/sampling": {"bytes": 8355, "sha256": "a80299c8dfc30739d05a0274526c361cb406c78a1ab5e923fcdf78d8c78ea73a"},- "judge:asset_v1:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/system": {"bytes": 8837, "sha256": "6dbacdd03ad0deab9f9264199b854eb37d8c29b58b5eaab8d1e78544f319a4f5"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3762, "sha256": "44cfcf9a5d4f9e2ccb999185652b82d5b75a589f088dedcce128cbe2e0d0b958"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4937, "sha256": "407f918497c4e96ed25d4a9c5c968993644c2a43fadfcc1bc6a0c4f7e6a9d99f"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/input": {"bytes": 3824, "sha256": "d2b8392354ef8ee0eaaa261eedb39f1c73fddc4373960c48d513e50b707daa57"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/messages": {"bytes": 3099, "sha256": "eee511c765a105054417cb3fa37bf67c24bcf9022da44dc086385878a61d5f67"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/sampling": {"bytes": 8355, "sha256": "a80299c8dfc30739d05a0274526c361cb406c78a1ab5e923fcdf78d8c78ea73a"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate/system": {"bytes": 8834, "sha256": "8b52aa8a0e80ff6ba98f494c36013745f5dd07a4e1040d72ff7f55aeeb907ba5"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/input": {"bytes": 3844, "sha256": "f53c37a588a498c52a403ce9178c0aea6e599a0b8ec944c1a9a88baf0e6ef175"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/messages": {"bytes": 3099, "sha256": "eee511c765a105054417cb3fa37bf67c24bcf9022da44dc086385878a61d5f67"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/sampling": {"bytes": 8355, "sha256": "a80299c8dfc30739d05a0274526c361cb406c78a1ab5e923fcdf78d8c78ea73a"},+ "judge:asset:runs/run/_assets/crate/artifacts/renders/r00/assets/crate_fix/system": {"bytes": 8834, "sha256": "8b52aa8a0e80ff6ba98f494c36013745f5dd07a4e1040d72ff7f55aeeb907ba5"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3762, "sha256": "44cfcf9a5d4f9e2ccb999185652b82d5b75a589f088dedcce128cbe2e0d0b958"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4937, "sha256": "407f918497c4e96ed25d4a9c5c968993644c2a43fadfcc1bc6a0c4f7e6a9d99f"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},"skills:runs/run:env@r00/routed": {"bytes": 164, "sha256": "eb0e2aca5a76b26e49a2d2c862a9405ef2e3e1130e719cca9a1f23fe1b264402"},"skills:runs/run:zone@r00/routed": {"bytes": 161, "sha256": "c2efc99419ed923cc9093c675fa27abc39d11613b9d2ac59e2cc5aea67d7d191"}@@ -585,10 +585,10 @@"job:zones_quay_water@r00/prompt": {"bytes": 52823, "sha256": "2a2e8d28955755da2036d29d9deb83a462d43e649e2e0845f0bab8cb746edf1c"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "5b1def419a1370ab2d9f2345bc4a5232ef3c793d4e4f1e01bd5ab9a3a8e88a7d"},"job:zones_quay_water@r00/system": {"bytes": 5614, "sha256": "b23b5fa6365b226f94f9cf4e9f3f7235628d22c1bee3ca22c6f44406ce518cdb"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "c1ace02589db83a171c73e42785af63acd95ac141b2c40cb1439e12ded9bd12c"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "c1ace02589db83a171c73e42785af63acd95ac141b2c40cb1439e12ded9bd12c"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},"skills:runs/run:env@r00/routed": {"bytes": 164, "sha256": "eb0e2aca5a76b26e49a2d2c862a9405ef2e3e1130e719cca9a1f23fe1b264402"},"skills:runs/run:zone@r00/routed": {"bytes": 161, "sha256": "c2efc99419ed923cc9093c675fa27abc39d11613b9d2ac59e2cc5aea67d7d191"}@@ -692,14 +692,14 @@"job:zones_quay_water@r00/prompt": {"bytes": 54501, "sha256": "daf87c38db35731f599f42c8028f4c1111a2d346ee7ac3c482f809208db9513a"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "5b1def419a1370ab2d9f2345bc4a5232ef3c793d4e4f1e01bd5ab9a3a8e88a7d"},"job:zones_quay_water@r00/system": {"bytes": 5614, "sha256": "b23b5fa6365b226f94f9cf4e9f3f7235628d22c1bee3ca22c6f44406ce518cdb"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "c1ace02589db83a171c73e42785af63acd95ac141b2c40cb1439e12ded9bd12c"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},- "judge:scene_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r01/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "c1ace02589db83a171c73e42785af63acd95ac141b2c40cb1439e12ded9bd12c"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},+ "judge:scene:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},+ "judge:scene:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r01/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},"skills:runs/run:env@r00/routed": {"bytes": 164, "sha256": "eb0e2aca5a76b26e49a2d2c862a9405ef2e3e1130e719cca9a1f23fe1b264402"},"skills:runs/run:refine@r01/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},@@ -749,14 +749,14 @@"job:zones_quay_water@r00/prompt": {"bytes": 53496, "sha256": "82b80505d5bb0ecc0ed5c7bc53f7d0bafdb8dc8102cb51c314e6628b3e7ffbcc"},"job:zones_quay_water@r00/scope": {"bytes": 379, "sha256": "5b1def419a1370ab2d9f2345bc4a5232ef3c793d4e4f1e01bd5ab9a3a8e88a7d"},"job:zones_quay_water@r00/system": {"bytes": 5614, "sha256": "b23b5fa6365b226f94f9cf4e9f3f7235628d22c1bee3ca22c6f44406ce518cdb"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "7fa5ae748295e277f0edb9fc51a241ac20ccf6c64a88431cb95bf08313fb661a"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},- "judge:scene_v2:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},- "judge:scene_v2:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},- "judge:scene_v2:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r01/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3755, "sha256": "7fa5ae748295e277f0edb9fc51a241ac20ccf6c64a88431cb95bf08313fb661a"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},+ "judge:scene:runs/run/artifacts/renders/r01/input": {"bytes": 4755, "sha256": "b386a92b7164f283a376ee84122a5ab109a1b709e28ade5a269337fe8c587ae0"},+ "judge:scene:runs/run/artifacts/renders/r01/messages": {"bytes": 5298, "sha256": "c9a488e3385bb803ce084eda9a1ad2ed58c338201f1ad6d7f416cf0070f1a9f1"},+ "judge:scene:runs/run/artifacts/renders/r01/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r01/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},"skills:runs/run:env@r00/routed": {"bytes": 164, "sha256": "eb0e2aca5a76b26e49a2d2c862a9405ef2e3e1130e719cca9a1f23fe1b264402"},"skills:runs/run:refine@r01/routed": {"bytes": 156, "sha256": "cae5ab46c39f9f4efb7ae3c4758b05d768025f93d525d5e13812b916a583fd30"},@@ -810,10 +810,10 @@"chat:planner:zone-layout#2/messages": {"bytes": 485, "sha256": "67e917190ee227c5ed3415ab4b03c072af313d7040704e7de671e63c42e2ab52"},"chat:planner:zone-layout#2/sampling": {"bytes": 3397, "sha256": "160445a83ede28afc0191547fd42d9341e694ef70aaad380bebf60ead0a65c49"},"chat:planner:zone-layout#2/system": {"bytes": 3326, "sha256": "494acec302dd23fa2fc6ee1eb6d48e873d6cf0ad98bad413e27578cf062313d4"},- "judge:scene_v2:runs/run/artifacts/renders/r00/input": {"bytes": 3772, "sha256": "b905874db5d6112f7b8a77387b6158354f02e473d901f29516e684e3d56b80ab"},- "judge:scene_v2:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},- "judge:scene_v2:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},- "judge:scene_v2:runs/run/artifacts/renders/r00/system": {"bytes": 21575, "sha256": "5e41f4637a72b54b410e31a08f362357d645437afaae85d69e15b7a5738aa307"},+ "judge:scene:runs/run/artifacts/renders/r00/input": {"bytes": 3772, "sha256": "b905874db5d6112f7b8a77387b6158354f02e473d901f29516e684e3d56b80ab"},+ "judge:scene:runs/run/artifacts/renders/r00/messages": {"bytes": 4930, "sha256": "562ef4bfc6b24c34604482cb73a02fb4b01da5c0e70685e3a1a4e1dda616a827"},+ "judge:scene:runs/run/artifacts/renders/r00/sampling": {"bytes": 14066, "sha256": "1f67b211de491259bed48f1499b8dbd9fa2037a903a637ba53befda40ade693d"},+ "judge:scene:runs/run/artifacts/renders/r00/system": {"bytes": 21572, "sha256": "eee5e34a58a91720755be68ed4d39cf127c634a484a4948e77c77ecce7e8cd16"},"skills:runs/run:baseline@r00/routed": {"bytes": 169, "sha256": "6ccf83dd133ef358e077dc682854e631d69bb44f6a3db6e91910c6f89d8d9455"},"skills:runs/run:env@r00/routed": {"bytes": 177, "sha256": "997130b7576c4ce74eaf0489faf7318c0c1815f8f5337cdc081b1b20295b8bd9"},"skills:runs/run:refine@r01/routed": {"bytes": 169, "sha256": "6ccf83dd133ef358e077dc682854e631d69bb44f6a3db6e91910c6f89d8d9455"},@@ -849,22 +849,22 @@"job:refine@r01/prompt": {"bytes": 7333, "sha256": "d0c65719f1734b9f7ee2f082d1b772314daebebdbafde9fdb36694833aed20c1"},"job:refine@r01/scope": {"bytes": 426, "sha256": "3a39df2f37665e52ae78549431ea567fa9591605332e152cd7d41d2b58914893"},"job:refine@r01/system": {"bytes": 2865, "sha256": "e8bdde0b2251c9f422c9a02c8c457543798280f5215e740312ffb73f1d0d867f"},- "judge:static_object_v1:runs/run/_cand/c0/artifacts/renders/r00/quick/input": {"bytes": 5545, "sha256": "6041286d7536f146c7f66b22021c7fcf38a73e1627be58ab10dd741251d9a242"},- "judge:static_object_v1:runs/run/_cand/c0/artifacts/renders/r00/quick/messages": {"bytes": 4246, "sha256": "029c3c03fe1f78f6f0d8b6b1b6bb4825d2851b55595a3c15c1a03f052b0b9771"},- "judge:static_object_v1:runs/run/_cand/c0/artifacts/renders/r00/quick/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/_cand/c0/artifacts/renders/r00/quick/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/_cand/c1/artifacts/renders/r00/quick/input": {"bytes": 5545, "sha256": "cec8a9d579957e61327fef43f3d0a0568555b0b162ffea5e140d78a6f1714524"},- "judge:static_object_v1:runs/run/_cand/c1/artifacts/renders/r00/quick/messages": {"bytes": 4246, "sha256": "029c3c03fe1f78f6f0d8b6b1b6bb4825d2851b55595a3c15c1a03f052b0b9771"},- "judge:static_object_v1:runs/run/_cand/c1/artifacts/renders/r00/quick/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/_cand/c1/artifacts/renders/r00/quick/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9108, "sha256": "060f49684b12ce285d2616a38f07d5c1b01c6c0bf15ccc9408a74f5ab7fcaba0"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "7df542d34f969f456ab2fcd720b5ccf3a4547b1c57ed59a81df0a84a4d540b9a"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10321, "sha256": "8d9c8fd2cf83b0c69df1888ac52b8021dec2b05af26ed4575ecf3d92d2d3c583"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "a800f572433fa02006cc5a02c883359497949586963a0edeb799e047048d8f2c"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/_cand/c0/artifacts/renders/r00/quick/input": {"bytes": 5545, "sha256": "6041286d7536f146c7f66b22021c7fcf38a73e1627be58ab10dd741251d9a242"},+ "judge:static_object:runs/run/_cand/c0/artifacts/renders/r00/quick/messages": {"bytes": 4246, "sha256": "029c3c03fe1f78f6f0d8b6b1b6bb4825d2851b55595a3c15c1a03f052b0b9771"},+ "judge:static_object:runs/run/_cand/c0/artifacts/renders/r00/quick/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/_cand/c0/artifacts/renders/r00/quick/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/_cand/c1/artifacts/renders/r00/quick/input": {"bytes": 5545, "sha256": "cec8a9d579957e61327fef43f3d0a0568555b0b162ffea5e140d78a6f1714524"},+ "judge:static_object:runs/run/_cand/c1/artifacts/renders/r00/quick/messages": {"bytes": 4246, "sha256": "029c3c03fe1f78f6f0d8b6b1b6bb4825d2851b55595a3c15c1a03f052b0b9771"},+ "judge:static_object:runs/run/_cand/c1/artifacts/renders/r00/quick/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/_cand/c1/artifacts/renders/r00/quick/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r00/input": {"bytes": 9108, "sha256": "060f49684b12ce285d2616a38f07d5c1b01c6c0bf15ccc9408a74f5ab7fcaba0"},+ "judge:static_object:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "7df542d34f969f456ab2fcd720b5ccf3a4547b1c57ed59a81df0a84a4d540b9a"},+ "judge:static_object:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r00/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 10321, "sha256": "8d9c8fd2cf83b0c69df1888ac52b8021dec2b05af26ed4575ecf3d92d2d3c583"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "a800f572433fa02006cc5a02c883359497949586963a0edeb799e047048d8f2c"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run/_cand/c0:baseline@r00/routed": {"bytes": 133, "sha256": "53e382eba5cd3e9e5438941f437cd2148cca27c331d46ccbb73fb8b1b092a9dc"},"skills:runs/run/_cand/c1:baseline@r00/routed": {"bytes": 133, "sha256": "53e382eba5cd3e9e5438941f437cd2148cca27c331d46ccbb73fb8b1b092a9dc"},"skills:runs/run:baseline@r00/routed": {"bytes": 133, "sha256": "53e382eba5cd3e9e5438941f437cd2148cca27c331d46ccbb73fb8b1b092a9dc"},@@ -899,14 +899,14 @@"job:refine@r02/prompt": {"bytes": 7510, "sha256": "da99d8d8621d2de04fbd11e07303ba2ffe71b771338f95d5b2bf705e2002c682"},"job:refine@r02/scope": {"bytes": 350, "sha256": "4e7e7adc88c67e32725d4fbf7d2f95b24d0166d57881f69cf543ab1c6936ea35"},"job:refine@r02/system": {"bytes": 2205, "sha256": "2ca6cc8997115b5940dcef874355b65c82f25ee8000b964fe0ad96dce25ac0e8"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 9513, "sha256": "397a05e8b108eb96ce4df58b830bb9bfabf0d38468b77243008f0c43d12cd594"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 6808, "sha256": "afe220b9286ff63c1e78b62cf853ead93c525947a873c6eaea80d9094a9d483e"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/input": {"bytes": 10722, "sha256": "7b7dcb7f4afcb9e67f86062daa7ab7dd547a9fb54fae165c8a032c84c40cebaf"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/messages": {"bytes": 7233, "sha256": "d6a82300f22e09f4cc9f7d5a09ec89e5c6e9b300eab6b8899fc61648ac406c21"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 9513, "sha256": "397a05e8b108eb96ce4df58b830bb9bfabf0d38468b77243008f0c43d12cd594"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 6808, "sha256": "afe220b9286ff63c1e78b62cf853ead93c525947a873c6eaea80d9094a9d483e"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r02/input": {"bytes": 10722, "sha256": "7b7dcb7f4afcb9e67f86062daa7ab7dd547a9fb54fae165c8a032c84c40cebaf"},+ "judge:static_object:runs/run/artifacts/renders/r02/messages": {"bytes": 7233, "sha256": "d6a82300f22e09f4cc9f7d5a09ec89e5c6e9b300eab6b8899fc61648ac406c21"},+ "judge:static_object:runs/run/artifacts/renders/r02/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r02/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run:baseline@r00/routed": {"bytes": 134, "sha256": "45a5c9d0f2d8993a70bbe476733dab64bd31eb2555d5850e2936724b07404966"},"skills:runs/run:refine@r01/routed": {"bytes": 134, "sha256": "45a5c9d0f2d8993a70bbe476733dab64bd31eb2555d5850e2936724b07404966"},"skills:runs/run:refine@r02/routed": {"bytes": 134, "sha256": "f11f63910f028add64068034f65c2875924ea57ea94bc407859b8b1e1050c6b6"}@@ -934,18 +934,18 @@"job:refine@r02/prompt": {"bytes": 7469, "sha256": "6670e07ec1dfbd701d3777c92d02e55a317bdd71c2e52e615973192828466a27"},"job:refine@r02/scope": {"bytes": 428, "sha256": "57e2f2633c63ce908a2842199385fe002439dcc7d6f68ffac0261d5be22cb8e1"},"job:refine@r02/system": {"bytes": 3208, "sha256": "e24e99f321231e5587cae7f4fbbbcc50fbd254768f4f4c860c4dfa8f83804810"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9307, "sha256": "1bab6f0b46195a81bcb86db47342c74c5a55d095435d24e05122a952253c3057"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 6756, "sha256": "447c252713a4d26c70c0b368c30217605f33543c3a329b6c32bee7099bdb1860"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10520, "sha256": "bada44f0350738b015ac2a5dfea12eb03fc89080f48ec848c793e87964d3fea8"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7181, "sha256": "0ee41f64bf25c8490c05f3478f11afde83f2d8fc7f4753553d7497f1ef83543a"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/input": {"bytes": 10516, "sha256": "e52a0b8b6bf1f9df4e18cd6289b2d80e98d4ae669cc0273e2e34f0ac7653a43b"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/messages": {"bytes": 7181, "sha256": "61b52dd59cc941e842e0e789faf595cd9ed24187eba10e7cc399e1ca792eb099"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r02/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/artifacts/renders/r00/input": {"bytes": 9307, "sha256": "1bab6f0b46195a81bcb86db47342c74c5a55d095435d24e05122a952253c3057"},+ "judge:static_object:runs/run/artifacts/renders/r00/messages": {"bytes": 6756, "sha256": "447c252713a4d26c70c0b368c30217605f33543c3a329b6c32bee7099bdb1860"},+ "judge:static_object:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r00/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 10520, "sha256": "bada44f0350738b015ac2a5dfea12eb03fc89080f48ec848c793e87964d3fea8"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 7181, "sha256": "0ee41f64bf25c8490c05f3478f11afde83f2d8fc7f4753553d7497f1ef83543a"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r02/input": {"bytes": 10516, "sha256": "e52a0b8b6bf1f9df4e18cd6289b2d80e98d4ae669cc0273e2e34f0ac7653a43b"},+ "judge:static_object:runs/run/artifacts/renders/r02/messages": {"bytes": 7181, "sha256": "61b52dd59cc941e842e0e789faf595cd9ed24187eba10e7cc399e1ca792eb099"},+ "judge:static_object:runs/run/artifacts/renders/r02/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r02/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run:baseline@r00/routed": {"bytes": 133, "sha256": "537d8d98029fd6345c9d4a73eaf3615c7924c18a405b3ab633283694e23b1804"},"skills:runs/run:refine@r01/routed": {"bytes": 133, "sha256": "518f7b7ba0343e508a61d4b00f2c7628ee771ef0cd1766d631194f988152e858"},"skills:runs/run:refine@r02/routed": {"bytes": 133, "sha256": "518f7b7ba0343e508a61d4b00f2c7628ee771ef0cd1766d631194f988152e858"}@@ -970,14 +970,14 @@"job:refine@r01/prompt": {"bytes": 9156, "sha256": "f8e4ad5d4b7d1ccdc149f77b7e6302e6d364dc62ccb9e98a45fb8c190f3c7f7d"},"job:refine@r01/scope": {"bytes": 644, "sha256": "f5e0c8fc1e3c0527734f358ad11929cedc505d872e61bbbfa4926d2cf5f4a1d0"},"job:refine@r01/system": {"bytes": 3208, "sha256": "e24e99f321231e5587cae7f4fbbbcc50fbd254768f4f4c860c4dfa8f83804810"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/input": {"bytes": 11212, "sha256": "054fc9378fe72d6263cc8be55114c6ffcd74e380ac652375c8873c708e632c6b"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 7002, "sha256": "90d9f0c3e869c20608b23834b3a93df8242b1dd3324fdfda3b91de9542337eb0"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11134, "sha256": "f825a8da4e119120216c23d2d3433cbf49ab7107bb97d58f7cd8359d7bd8eb45"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 12397, "sha256": "942ff240dcba105727d8ff15dbc2e0b0535ae4958ec8731ea993d70d79b49bdc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7429, "sha256": "70fca81684fcce477eadc3e0d5c329b3dcc31dcc527958abf29e484a0d555a7c"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11134, "sha256": "f825a8da4e119120216c23d2d3433cbf49ab7107bb97d58f7cd8359d7bd8eb45"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/artifacts/renders/r00/input": {"bytes": 11212, "sha256": "054fc9378fe72d6263cc8be55114c6ffcd74e380ac652375c8873c708e632c6b"},+ "judge:static_object:runs/run/artifacts/renders/r00/messages": {"bytes": 7002, "sha256": "90d9f0c3e869c20608b23834b3a93df8242b1dd3324fdfda3b91de9542337eb0"},+ "judge:static_object:runs/run/artifacts/renders/r00/sampling": {"bytes": 11134, "sha256": "f825a8da4e119120216c23d2d3433cbf49ab7107bb97d58f7cd8359d7bd8eb45"},+ "judge:static_object:runs/run/artifacts/renders/r00/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 12397, "sha256": "942ff240dcba105727d8ff15dbc2e0b0535ae4958ec8731ea993d70d79b49bdc"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 7429, "sha256": "70fca81684fcce477eadc3e0d5c329b3dcc31dcc527958abf29e484a0d555a7c"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11134, "sha256": "f825a8da4e119120216c23d2d3433cbf49ab7107bb97d58f7cd8359d7bd8eb45"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run:baseline@r00/routed": {"bytes": 104, "sha256": "754997d8db8e89f9b7b0379fceda29998c2e42d362fcaec169d7bd48fc097d3b"},"skills:runs/run:refine@r01/routed": {"bytes": 104, "sha256": "754997d8db8e89f9b7b0379fceda29998c2e42d362fcaec169d7bd48fc097d3b"}},@@ -1001,14 +1001,14 @@"job:refine@r01/prompt": {"bytes": 7307, "sha256": "51c6293407b6d9352ac4e6c529907570d21b681e740f2f548c359f2d061c499c"},"job:refine@r01/scope": {"bytes": 428, "sha256": "186e31cec33c5144bbc9713eb76374ef7608534682433c9a9288191efc417ec4"},"job:refine@r01/system": {"bytes": 3208, "sha256": "e24e99f321231e5587cae7f4fbbbcc50fbd254768f4f4c860c4dfa8f83804810"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9108, "sha256": "de6324c2e5fffad8f4ae1e87cb2e54e6c347b1f9cc8d0f07479274fd6acb4e31"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "6e3964ce3478de892b3c5dde9eb3154e9a54c2a669315a8e6f4b808c15dd930c"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10321, "sha256": "0bdbcf6b262451c2f914db0f1501ffa30d50795fc7f8c6dbea0fdaf8f1ed2d14"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "0e3a5d3ed25799eec213d2d7fd92ae5009e4d88d5a818eeaf0aef3de8d47b932"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/artifacts/renders/r00/input": {"bytes": 9108, "sha256": "de6324c2e5fffad8f4ae1e87cb2e54e6c347b1f9cc8d0f07479274fd6acb4e31"},+ "judge:static_object:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "6e3964ce3478de892b3c5dde9eb3154e9a54c2a669315a8e6f4b808c15dd930c"},+ "judge:static_object:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r00/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 10321, "sha256": "0bdbcf6b262451c2f914db0f1501ffa30d50795fc7f8c6dbea0fdaf8f1ed2d14"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "0e3a5d3ed25799eec213d2d7fd92ae5009e4d88d5a818eeaf0aef3de8d47b932"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run:baseline@r00/routed": {"bytes": 133, "sha256": "537d8d98029fd6345c9d4a73eaf3615c7924c18a405b3ab633283694e23b1804"},"skills:runs/run:refine@r01/routed": {"bytes": 133, "sha256": "537d8d98029fd6345c9d4a73eaf3615c7924c18a405b3ab633283694e23b1804"}},@@ -1034,14 +1034,14 @@"chat:planner:planner/messages": {"bytes": 1354, "sha256": "5b1a589e27562a3668649aed4be4857a34804a2acdc140eb8fb3630d970f71a6"},"chat:planner:planner/sampling": {"bytes": 6272, "sha256": "04dbfbbd42d8b93a12571a1908beacc7eb471e6f2c0fc077371aa86e2a99fb40"},"chat:planner:planner/system": {"bytes": 18298, "sha256": "0483e07168e1b18aeb18cda38887e719a4c3993161c600c61da469ff2fccd4f4"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/input": {"bytes": 9125, "sha256": "cb85607f665756a289d7f4070a59cd5febf398ef289ec123c6060a3313d9c011"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "6e3964ce3478de892b3c5dde9eb3154e9a54c2a669315a8e6f4b808c15dd930c"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r00/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/input": {"bytes": 10338, "sha256": "a5ee63d3b5d902e5e111abce6463ff9a3fe1a1af21a3c6a9d284b76137b64b49"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "0e3a5d3ed25799eec213d2d7fd92ae5009e4d88d5a818eeaf0aef3de8d47b932"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},- "judge:static_object_v1:runs/run/artifacts/renders/r01/system": {"bytes": 15181, "sha256": "3575c0106dc2f760c7b997e75932b9e37662f45394bba5516c5feebac72fe1bc"},+ "judge:static_object:runs/run/artifacts/renders/r00/input": {"bytes": 9125, "sha256": "cb85607f665756a289d7f4070a59cd5febf398ef289ec123c6060a3313d9c011"},+ "judge:static_object:runs/run/artifacts/renders/r00/messages": {"bytes": 6699, "sha256": "6e3964ce3478de892b3c5dde9eb3154e9a54c2a669315a8e6f4b808c15dd930c"},+ "judge:static_object:runs/run/artifacts/renders/r00/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r00/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},+ "judge:static_object:runs/run/artifacts/renders/r01/input": {"bytes": 10338, "sha256": "a5ee63d3b5d902e5e111abce6463ff9a3fe1a1af21a3c6a9d284b76137b64b49"},+ "judge:static_object:runs/run/artifacts/renders/r01/messages": {"bytes": 7124, "sha256": "0e3a5d3ed25799eec213d2d7fd92ae5009e4d88d5a818eeaf0aef3de8d47b932"},+ "judge:static_object:runs/run/artifacts/renders/r01/sampling": {"bytes": 11368, "sha256": "d689cdebe9d63a438597fd51e16228f8d5fed5d5693c736223ce782f62a12242"},+ "judge:static_object:runs/run/artifacts/renders/r01/system": {"bytes": 15178, "sha256": "1becebca9df2996a380a8bedede59de191c58f5021c6f687a3dfd5abc673c4bc"},"skills:runs/run:baseline@r00/routed": {"bytes": 148, "sha256": "67ba73ea33f42a757e533e24b318438e321943eeb47470a930fbcfda97e72e93"},"skills:runs/run:refine@r01/routed": {"bytes": 148, "sha256": "67ba73ea33f42a757e533e24b318438e321943eeb47470a930fbcfda97e72e93"}},diff --git a/harness/tests/reference/test_judge_diff.py b/harness/tests/reference/test_judge_diff.pyindex 80aedb66..5990b973 100644--- a/harness/tests/reference/test_judge_diff.py+++ b/harness/tests/reference/test_judge_diff.py@@ -20,7 +20,7 @@ from tests.judges.conftest import (make_spec,)-REF = load_rubric("reference_v1")+REF = load_rubric("reference")DIFF = {"mismatches": [{"kind": "wrong_part_count", "target": "Legs","detail": "reference shows 4 legs, render shows 2", "severity": "critical"}],@@ -60,7 +60,7 @@ def test_a_failing_diff_never_breaks_the_verdict(ref_input, cache_dir):model = FakeChatModel(by_label={"reference_diff": [{"garbage": 1}],"judge": [good_reply(REF, ["A1", "A2"], 0.8)]})j = _judge(model, cache_dir=cache_dir).judge(ref_input)- assert j.overall > 0 and j.rubric == "reference_v1"+ assert j.overall > 0 and j.rubric == "reference"@pytest.mark.parametrize(("value", "diffed"), [("off", False), ("OFF", False), ("of", True)])diff --git a/harness/tests/scene_gates/test_frame_metrics.py b/harness/tests/scene_gates/test_frame_metrics.pyindex 30d0a3c5..b7cfa62d 100644--- a/harness/tests/scene_gates/test_frame_metrics.py+++ b/harness/tests/scene_gates/test_frame_metrics.py@@ -136,7 +136,7 @@ def test_no_hint_prescribes_one_camera_or_light_construction():def test_scene_rubric_caps_fire_on_frame_kinds():- rubric = load_rubric("scene_v2")+ rubric = load_rubric("scene")rep = frame_findings(_metrics(_chk("Establishing", mean_lum=0.07, dark_frac=0.41, content_frac=0.1)))res = apply_caps(rubric, 0.9, [rep], {}, [])fired = {c.rule: c.cap for c in res.caps_applied}diff --git a/harness/tests/scene_gates/test_judge_views.py b/harness/tests/scene_gates/test_judge_views.pyindex b2c26bb9..dc83fd03 100644--- a/harness/tests/scene_gates/test_judge_views.py+++ b/harness/tests/scene_gates/test_judge_views.py@@ -21,7 +21,7 @@ def test_default_selection_is_authored_plus_overviews():rs = _rs()assert len(rs.views) == 2 * (3 + 7)sel = select_judge_views(rs)- # 3 authored + the 4 overview tiles (scene_v2: the contact camera is the fourth) + 2 at t=1.5+ # 3 authored + the 4 overview tiles (scene: the contact camera is the fourth) + 2 at t=1.5assert len(sel.views) == 9 <= JUDGE_MAX_VIEWS == 10keys = [(v.name, v.time_s) for v in sel.views]# authored t=0, the overview rig t=0, first two authored at t=1.5@@ -40,7 +40,7 @@ def test_default_selection_is_authored_plus_overviews():def test_the_rig_has_four_judged_overviews_and_the_contact_camera_is_the_fourth():- """scene_v2: ``overview_low_front`` (az 20°, el 7°) is in the rig, judged, and fitted as the+ """scene: ``overview_low_front`` (az 20°, el 7°) is in the rig, judged, and fitted as thecontact camera (orbit.mjs takes elevation < 10° to fitContactCamera); the eye rig stays diagnostic."""from codeverse3d.spatial.render_scene import MAX_JUDGE_OVERVIEWSdiff --git a/harness/tests/scene_gates/test_live_frames.py b/harness/tests/scene_gates/test_live_frames.pyindex b2002e17..f6ee8315 100644--- a/harness/tests/scene_gates/test_live_frames.py+++ b/harness/tests/scene_gates/test_live_frames.py@@ -75,7 +75,7 @@ def test_example_scene_passes_frame_gate_and_judge_subset(starter_ws: Workspace)assert rs.out_dir == str(out)assert all(v.judge is not None for v in rs.views)stamped = {(v.name, v.time_s) for v in rs.views if v.judge}- # 3 authored at t=0 + the 4 overview-rig tiles (scene_v2: the contact camera is the fourth)+ # 3 authored at t=0 + the 4 overview-rig tiles (scene: the contact camera is the fourth)# + the first two authored at t=1.5; the harness's eye-level rig is a diagnostic and never# reaches the judgeassert len(stamped) == 9 <= JUDGE_MAX_VIEWSdiff --git a/harness/tests/scene_prompts/test_quality_teaching.py b/harness/tests/scene_prompts/test_quality_teaching.pyindex 13012070..e531dc0a 100644--- a/harness/tests/scene_prompts/test_quality_teaching.py+++ b/harness/tests/scene_prompts/test_quality_teaching.py@@ -12,14 +12,14 @@ COMPOSITION_DEFECTS = ("flat_ground", "monotonous_vegetation", "empty_midground"def test_scene_rubric_craft_defects_inform_and_cannot_zero_a_scene() -> None:- r = load_rubric("scene_v2")+ r = load_rubric("scene")assert set(COMPOSITION_DEFECTS) <= {d.id for d in r.defects}for did in COMPOSITION_DEFECTS:d = r.defect(did)assert 0.0 < d.penalty <= 0.05 and d.cap is None, did# the five craft items together take at most 0.20: they inform, they cannot sink a sceneassert sum(r.defect(did).penalty for did in COMPOSITION_DEFECTS) == pytest.approx(0.18)- # scene_v2 (2026-10-14): the whole checklist sums past 1.0 (was 0.60 under v1's 0.62 bound). The+ # scene (2026-10-14): the whole checklist sums past 1.0 (was 0.60 under v1's 0.62 bound). The# weight that moved is the placement / detail items — floating 0.10, sunken 0.08, interpenetration# 0.08, wrong_scale 0.06, missing_planned_element 0.06, hero_misplaced 0.06, hard-edged particles# 0.04 — the same weights the object rubric gives floating_part / interpenetration, and each ofdiff --git a/harness/tests/scene_runtime/test_orbit.py b/harness/tests/scene_runtime/test_orbit.pyindex c7bc2dc0..68151074 100644--- a/harness/tests/scene_runtime/test_orbit.py+++ b/harness/tests/scene_runtime/test_orbit.py@@ -73,7 +73,7 @@ console.log(JSON.stringify({{def test_contact_camera_stands_at_ground_level_outside_the_content(tmp_path):- """scene_v2's ``overview_low_front`` (elevation < 10°): 1.2× the content span from the centre,+ """scene's ``overview_low_front`` (elevation < 10°): 1.2× the content span from the centre,0.4 m above the ground, aimed at the ground edge of the content box; fog kept."""body = f"""import {{ fitOrbitCameras, fitContactCamera }} from './lib/orbit.mjs';diff --git a/harness/tests/scene_runtime/test_render_blender.py b/harness/tests/scene_runtime/test_render_blender.pyindex d0c1adf3..5eecc7c6 100644--- a/harness/tests/scene_runtime/test_render_blender.py+++ b/harness/tests/scene_runtime/test_render_blender.py@@ -73,7 +73,7 @@ def test_a_fixture_scene_renders_two_cameras_at_two_times(moving):for v in rs.views:with Image.open(v.path) as im:assert im.size == (W, H)- # the four overviews (scene_v2: the contact camera overview_low_front is the fourth) get pixels+ # the four overviews (scene: the contact camera overview_low_front is the fourth) get pixels# at t0 only; the eye-level rig is measured, never renderedrig = sorted((v.name, v.time_s) for v in rs.views if v.name.startswith(("overview", "eye")))assert rig == [("overview_back_left", 0.0), ("overview_front_right", 0.0), ("overview_low_front", 0.0), ("overview_top", 0.0)]diff --git a/harness/tests/texturing/conftest.py b/harness/tests/texturing/conftest.pyindex 6bccc127..0c99a5d7 100644--- a/harness/tests/texturing/conftest.py+++ b/harness/tests/texturing/conftest.py@@ -107,7 +107,7 @@ class FakeJudge:overall, scores = self.verdicts[min(self.calls, len(self.verdicts) - 1)]self.calls += 1summary = "judge_error: boom" if self.degraded else "ok"- return Judgment(rubric="static_object_v1", scores=scores, overall=overall, passed=overall >= 0.72, summary=summary,+ return Judgment(rubric="static_object", scores=scores, overall=overall, passed=overall >= 0.72, summary=summary,usage=Usage(cost_usd=0.01))diff --git a/harness/tests/texturing/test_gate.py b/harness/tests/texturing/test_gate.pyindex 91520789..b26ab90a 100644--- a/harness/tests/texturing/test_gate.py+++ b/harness/tests/texturing/test_gate.py@@ -16,8 +16,8 @@ def _asset(tid: str, seam: float, ok_path: Path) -> TextureAsset:@pytest.mark.parametrize(("rubric", "criterion"), [- ("static_object_v1", "materials"), ("reference_v1", "material_color"), ("asset_v1", "material_truth"),- ("scene_v2", "materials_shaders_effects"), ("articulated_v1", ""), ("shader_v2", ""),+ ("static_object", "materials"), ("reference", "material_color"), ("asset", "material_truth"),+ ("scene", "materials_shaders_effects"), ("articulated", ""), ("shader", ""),])def test_the_rubric_flags_its_materials_criterion(rubric, criterion):"""N64: the gate sniffed criterion ids by prefix (law 4); a rubric now says which one it is."""