List your company

Physics Problem Solving on PHYSICS (held-out)

20.8DBSEXPERT
5.3049.32713.3517.373Feb 11, 2026

Evaluation results

MethodLinks
EXPERTModel=Qwen3-1.7B optimized2026.0220.8—
DataChef-32BOracle Upper Bound=true, Samples=322026.0210.4—
Gemini-3-Pro2026.029.269.7
Kimi-K22026.02919.7
Qwen3-Next ⊕ Kimi-K2Reasoning backbone=Qwen3-Next-80B, Coding backbone=Kimi-K2-Instruct2026.028.948.7
DataChef-32B2026.028.761.4
SOURCEbest2026.028.5—
SOURCEavg2026.026.1—
Qwen3-32B2026.025.911