Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
tests: cite the two _decode_answers by symbol, not by line
Nandakishor702043c9e862
diff --git a/tests/test_confidence.py b/tests/test_confidence.pyindex c38730b5..d11dabf9 100644--- a/tests/test_confidence.py+++ b/tests/test_confidence.py@@ -433,8 +433,9 @@ check("export/laya re-exports the same tuple", laya.GATE_STATES, _confidence.GAT# --------------------------------- the two doc pages must attribute confidence per question type# `docs/structured.md` and `docs/questions-and-answers.md` each opened their confidence section with# "`confidence` is normalized entropy". Neither agent writes that for every type:-# `Agent._decode_answers` (`laya/agent.py:1369-1398`) and `ONNXAgent._decode_answers`-# (`laya/onnx_agent.py:715-785`) put `confidence_from_probs(p, k)` = `1 - H(p) / log(k)` in the+# `Agent._decode_answers` (`laya/agent.py::Agent._decode_answers`) and `ONNXAgent._decode_answers`+# (`laya/onnx_agent.py::ONNXAgent._decode_answers`) put `confidence_from_probs(p, k)` =+# `1 - H(p) / log(k)` in the# `choice` and `score` answers and `max(p_true, 1 - p_true)` in the `noul` one. Each page printed a# `noul` in the very block its sentence introduces, so the sentence was falsifiable from the page# alone: the Q&A page's sample answer is `{"type": "noul", "noul": 0.8727, "confidence": 0.8727}`,