List your company

Medical Diagnostic Reasoning on MedCaseReasoning official 897-case (test)

32.83AccuracyGRPO
20.079623.389826.730.0102Apr 13, 2026

Evaluation results

MethodLinks
GRPOCategory=RL, Backbone=LLaMA3.1-8B-Instruct2026.0432.8373.24
GRPOCategory=RL, Backbone=Qwen2.5-7B-Instruct2026.0432.8173.12
CGCLCategory=Ours, Backbone=LLaMA3.1-8B-Instruct2026.0432.6772.89
CGCLCategory=Ours, Backbone=Qwen2.5-7B-Instruct2026.0432.6372.84
DPOCategory=RL, Backbone=LLaMA3.1-8B-Instruct2026.0432.2971.22
DPOCategory=RL, Backbone=Qwen2.5-7B-Instruct2026.0432.2471.18
SFT-CoTCategory=SFT, Backbone=LLaMA3.1-8B-Instruct2026.0432.1460.86
SFT-CoTCategory=SFT, Backbone=Qwen2.5-7B-Instruct2026.0432.0860.79
SFT-GTCategory=SFT, Backbone=LLaMA3.1-8B-Instruct2026.0431.7963.81
SFT-GTCategory=SFT, Backbone=Qwen2.5-7B-Instruct2026.0431.7263.74
SFT-CGCL(Stage3)Category=SFT, Backbone=Qwen2.5-7B-Instruct2026.0430.1270.78
SFT-CGCL(Stage3)Category=SFT, Backbone=LLaMA3.1-8B-Instruct2026.0430.0770.92
CGCLCategory=Ours, Backbone=Qwen2.5-3B-Instruct2026.0428.9571.3
GRPOCategory=RL, Backbone=LLaMA3.2-3B-Instruct2026.0428.9571.3
GRPOCategory=RL, Backbone=Qwen2.5-3B-Instruct2026.0428.971.25
CGCLCategory=Ours, Backbone=LLaMA3.2-3B-Instruct2026.0428.971.35
DPOCategory=RL, Backbone=LLaMA3.2-3B-Instruct2026.0428.6369.31
DPOCategory=RL, Backbone=Qwen2.5-3B-Instruct2026.0428.5969.27
SFT-CoTCategory=SFT, Backbone=LLaMA3.2-3B-Instruct2026.0428.3359.91
SFT-CoTCategory=SFT, Backbone=Qwen2.5-3B-Instruct2026.0428.2759.83
SFT-GTCategory=SFT, Backbone=Qwen2.5-3B-Instruct2026.0428.1662.38
SFT-GTCategory=SFT, Backbone=LLaMA3.2-3B-Instruct2026.0428.1262.49
SFT-CGCL(Stage3)Category=SFT, Backbone=LLaMA3.2-3B-Instruct2026.0426.6468.59
SFT-CGCL(Stage3)Category=SFT, Backbone=Qwen2.5-3B-Instruct2026.0426.5868.67
Trust-Think & AnswerCategory=Prompt, Backbone=LLaMA3.1-8B-Instruct2026.0426.3865.47
Trust-Think & AnswerCategory=Prompt, Backbone=Qwen2.5-7B-Instruct2026.0426.3165.52
Think & AnswerCategory=Prompt, Backbone=LLaMA3.1-8B-Instruct2026.0425.7362.58
Think & AnswerCategory=Prompt, Backbone=Qwen2.5-7B-Instruct2026.0425.6862.43
Think & AnswerCategory=Prompt, Backbone=LLaMA3.2-3B-Instruct2026.0423.5161.29
Think & AnswerCategory=Prompt, Backbone=Qwen2.5-3B-Instruct2026.0423.4761.32
Trust-Think & AnswerCategory=Prompt, Backbone=LLaMA3.2-3B-Instruct2026.0423.4363.36
Trust-Think & AnswerCategory=Prompt, Backbone=Qwen2.5-3B-Instruct2026.0423.4263.28
VanillaCategory=Prompt, Backbone=LLaMA3.1-8B-Instruct2026.0422.46—
VanillaCategory=Prompt, Backbone=Qwen2.5-7B-Instruct2026.0422.37—
VanillaCategory=Prompt, Backbone=Qwen2.5-3B-Instruct2026.0420.61—
VanillaCategory=Prompt, Backbone=LLaMA3.2-3B-Instruct2026.0420.57—