List your company

Science Question Answering on GPQA D

67.4Accuracy (Avg@8)CERL full
55.02458.23761.4564.663May 22, 2026

Evaluation results

MethodLinks
CERL fullTraining=SFT+RL2026.0567.4—
HaloTraining=SFT+RL2026.0564.6—
Correctness-only RLTraining=RL2026.0564.6—
Long-answer SFTTraining=SFT2026.0562.8—
Fixed-ratio erasure + free answerTraining=None2026.0562.6—
Base Qwen3-8BTraining=N/A2026.0561.9—
Length-penalty RLTraining=RL2026.0561.3—
TokenSkipTraining=SFT2026.0555.5—
IOAStudent Model=Qwen2.5-7B, Teacher LLM=DeepSeek-R12026.02—16.23