Share your thoughts, 1 month free Claude Pro on usSee more

Scientific Reasoning on JEEBench

57.34Mean Accuracy

Reason, Reward, Refine

Updated 18d ago

Evaluation Results

Method	Links
Reason, Reward, Refine 2026.07		57.34
Reason, Reward, Refine 2026.07		54.86
Qwen2.5 32B + RL (synthetic) 2026.04		52.28
Reason, Reward, Refine 2026.07		50.23
Reason, Reward, Refine 2026.07		49.17
DPO 2026.07		41.65
RAG 2026.07		41.25
RAG 2026.07		40.83
RAG 2026.07		40
FT 2026.07		39.02
DPO 2026.07		38.21
FT 2026.07		38.04
CoT 2026.07		37.53
FT 2026.07		36.59
CoT 2026.07		35.9
DPO 2026.07		35.77
RAG 2026.07		35.5
Qwen2.5 32B 2026.04		34.38
CoT 2026.07		32.52
CoT 2026.07		30.21
DPO 2026.07		30.08
FT 2026.07		28.46