Medical Reasoning on MedQA

92.8Accuracy

ReConcile

Updated 25d ago

Evaluation Results

Method	Links
ReConcile 2025.08		92.8
Med-Gemini 2024.04		91.1
TMA-AllCompon 2025.08		90.7
MedAgents 2025.08		90.3
Previous SoTA (Nori et al. 2023) 2024.04		90.2
Best GPT-4 Method 2024.04		90.2
DeepSeek-V4 2026.05		88.92
MDAgents 2025.08		88.7
GPT-5.0 2026.05		87.43
DyLAN 2025.08		86.8
MedGuideX-9B 2026.05		80.13
Kimi-2.6 2026.05		79.26
MedGuideX-4B 2026.05		77.06
RL with CPG-Derived Process Rewards 2026.05		75.73
Qwen3.5-9B 2026.05		74.94
Qwen3.5-9B 2026.05		74.71
Qwen3.5-4B 2026.05		73.45
Qwen3.5-9B 2026.05		73.21
Hulu-Med-7B 2026.05		72.98
Qwen3.5-4B 2026.05		72.43
Huatuo-o1 2026.02		72.4
Ours 2026.06		72.3
Qwen3.5-4B 2026.05		71.41
HuatuoGPT-o1-8B 2026.05		71.09
MedReason-8B 2026.05		70.86
m1k 2026.06		70.4
Random 2026.06		69
Middle Perplexity 2026.06		68.6
S2L 2026.06		68.5
MedGemma-1.5-4B 2026.05		67.79
Embedding Diversity 2026.06		67.3
MedVerse 2026.02		66.4
Learnability 2026.06		65.9
Fine-tuning with CPG 2026.05		65.04
Llama-Aloe-Beta-8B 2026.05		64.65
Lingshu-7B 2026.05		64.49
MedReason 2026.02		63.9
TMA-AllCompon 2025.08		62.1
HIPPO 2026.06		61.7
Qwen2.5_7B 2026.06		61.2
SP3F 2026.06		61.1
CPGPrompt 2026.05		59.15
Original (LLaMA-3.1-8B-Instruct) 2026.02		58.7
MedVerse 2026.02		58.6
PRM RL 2026.06		57.7
PREF-GRPO 2026.06		56.6
Original (Qwen2.5-7B-Instruct) 2026.02		56.2
MedReason 2026.02		56.2
MediPhi-Instruct-4B 2026.05		53.34
Clinical-R1-3B 2026.05		52.55
Qwen2.5_7B 2026.06		49.5
Qwen2.5_7B 2026.06		49.4
DyLAN 2025.08		49.2
ReConcile 2025.08		48.5
Claude-Haiku-4.5 2026.05		48.08
MedAgents 2025.08		46.8
BioMistral-7B 2026.05		45.8
TMA-AllCompon 2025.08		45.5
Llava-Med-v1.5-7B 2026.05		44.78
MDAgents 2025.08		39.6
MedAlpaca-7B 2026.05		38.81