List your company

Code Generation on Code (test)

36.5AccuracyNon-hacking
14.55620.25325.9531.647Apr 17, 2026

Evaluation results

MethodLinks
Non-hackingBase Model=Qwen2.5-3B-Instruct2026.0436.5—
RFT+GRIFTBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=202026.0423.332.8
Starting-pointBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=202026.0419.649.5
No-interventionBase Model=Qwen2.5-3B-Instruct2026.0416.281.9
RFT+RandomBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=202026.0415.954.6
RFT+TraceBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=202026.0415.467.2