List your company

Medical Diagnosis and Reasoning on MEDFOUND (OOD)

29.83DxCode AccuracyCGCL
19.450822.145424.8427.5346Apr 13, 2026

Evaluation results

MethodLinks
CGCLCategory=Ours, Backbone=LLaMA3.2-3B-Instruct, Zero-shot=true2026.0429.8331.2471.36
CGCLBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=Ours2026.0429.2830.6470.89
GRPOBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=RL2026.0429.0130.3570.76
SFT-CGCL(Stage3)Category=SFT, Backbone=LLaMA3.2-3B-Instruct, Zero-shot=true2026.0428.4529.6868.21
GRPOCategory=RL, Backbone=LLaMA3.2-3B-Instruct, Zero-shot=true2026.0428.1729.4670.58
DPOBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=RL2026.0428.1529.568.9
Trust-Think & AnswerCategory=Prompt, Backbone=LLaMA3.2-3B-Instruct, Zero-shot=true2026.0427.928.9563.12
SFT-CGCL(Stage3)Backbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=SFT2026.0427.3528.6267.84
Trust-Think & AnswerBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=Prompt2026.0426.2427.5162.45
SFT-CoTBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=SFT2026.0425.126.263.88
SFT-GTBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=SFT2026.0424.3825.462.1
Think & AnswerBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=Prompt2026.0423.1224.260.15
VanillaBackbone=Qwen2.5-3B-Instruct, Zero-shot=true, Category=Prompt2026.0419.8520.558.24