Official Codebase of 3DCodeVerse
Python
Reorder highlights to present harness and dataset before results
gaoypengd8741e6fcfa2
diff --git a/README.md b/README.mdindex 5b641c6f..5fc09b05 100644--- a/README.md+++ b/README.md@@ -30,6 +30,30 @@ https://github.com/user-attachments/assets/71374368-ccf0-410b-b2ee-fac7f7405fbc## Highlights+### 3DCodeVerse Harness++<p align="center"><img src="assets/harness_overview.png" alt="Overview of the 3DCodeVerse Harness" width="95%"></p>++An LLM planner defines the target asset, and a DAG of sub-agents codes and assembles a 3D program.+Every round, the harness collects three kinds of feedback:++1. **Execution** — build and runtime logs.+2. **Measurement** — dimensions, contacts, collisions, kinematics and placement, reported per part+ with numerical correction hints.+3. **Vision** — a rubric-guided VLM judge on labelled multi-view renders.++The score is computed in code from rubric weights, defect penalties and measured caps, and the+measurements veto visual claims they rule out. The feedback becomes localized repair tasks for a fix+agent, and the best-scoring round is the deliverable. Because every round is recorded, including+successes, failures, repairs and preference pairs, the harness also serves as the corpus's data+engine.++### 3DCodeVerse-2M Dataset++<p align="center"><img src="assets/dataset_overview.jpg" alt="Curation and composition of 3DCodeVerse-2M" width="95%"></p>++### Results+**The harness lifts every backbone it runs.** On 100 text-to-3D object prompts, the same modelinside the 3DCodeVerse Harness beats its own native coding harness:@@ -55,26 +79,6 @@ on 805 fixed tasks:| Qwen3.8-27B | 51.9 | 74.8 | 65.5 | 80.0 | 57.0 | 72.5 || + SFT + DPO | **92.9** | **93.2** | **98.5** | **90.0** | **71.5** | **92.5** |-## How it works--<p align="center"><img src="assets/harness_overview.png" alt="Overview of the 3DCodeVerse Harness" width="95%"></p>--An LLM planner defines the target asset, and a DAG of sub-agents codes and assembles a 3D program.-Every round, the harness collects three kinds of feedback:--1. **Execution** — build and runtime logs.-2. **Measurement** — dimensions, contacts, collisions, kinematics and placement, reported per part- with numerical correction hints.-3. **Vision** — a rubric-guided VLM judge on labelled multi-view renders.--The score is computed in code from rubric weights, defect penalties and measured caps, and the-measurements veto visual claims they rule out. The feedback becomes localized repair tasks for a fix-agent, and the best-scoring round is the deliverable. Because every round is recorded, including-successes, failures, repairs and preference pairs, the harness also serves as the corpus's data-engine.--<p align="center"><img src="assets/dataset_overview.jpg" alt="Curation and composition of 3DCodeVerse-2M" width="95%"></p>-## Repository layout```