List your company
ResearchFormerly WizWand

Medical Reasoning on MedQA

92.8AccuracyReConcile
36.650451.227765.80580.3823Apr 29, 2024Sep 7, 2024Jan 17, 2025May 29, 2025Oct 7, 2025Feb 16, 2026Jun 28, 2026
Updated 4mo ago

Evaluation results

MethodLinks
ReConcileBase Model=GPT-4o2025.0892.8
Med-Gemini2024.0491.1
TMA-AllComponBase Model=GPT-4o2025.0890.7
MedAgentsBase Model=GPT-4o2025.0890.3
Previous SoTA (Nori et al. 2023)2024.0490.2
Best GPT-4 Method2024.0490.2
DeepSeek-V42026.0588.92
MDAgentsBase Model=GPT-4o2025.0888.7
GPT-5.02026.0587.43
DyLANBase Model=GPT-4o2025.0886.8
MedGuideX-9BBackbone=Qwen3.5-9B2026.0580.13
Kimi-2.62026.0579.26
MedGuideX-4BBackbone=Qwen3.5-4B2026.0577.06
RL with CPG-Derived Process RewardsBackbone=Qwen3.5-9B2026.0575.73
Qwen3.5-9BPrompting Strategy=3-Shot In-Context Learning2026.0574.94
Qwen3.5-9BPrompting Strategy=RAG with Guidelines2026.0574.71
Qwen3.5-4BPrompting Strategy=RAG with Guidelines2026.0573.45
Qwen3.5-9BPrompting Strategy=Zero-shot2026.0573.21
Hulu-Med-7B2026.0572.98
Qwen3.5-4BPrompting Strategy=Zero-shot2026.0572.43
Huatuo-o1Backbone=LLaMA-3.1-8B-Instruct2026.0272.4
OursModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0672.3
Qwen3.5-4BPrompting Strategy=3-Shot In-Context Learning2026.0571.41
HuatuoGPT-o1-8B2026.0571.09
MedReason-8B2026.0570.86
m1kModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0670.4
RandomModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0669
Middle PerplexityModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0668.6
S2LModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0668.5
MedGemma-1.5-4B2026.0567.79
Embedding DiversityModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0667.3
MedVerseBackbone=LLaMA-3.1-8B-Instruct2026.0266.4
LearnabilityModel=Qwen2.5-7B-Instruct, Training=Finetuning, Sample size=1000 examples2026.0665.9
Fine-tuning with CPGBackbone=Qwen3.5-9B2026.0565.04
Llama-Aloe-Beta-8B2026.0564.65
Lingshu-7B2026.0564.49
MedReasonBackbone=LLaMA-3.1-8B-Instruct2026.0263.9
TMA-AllComponBase Model=MedGemma-4B2025.0862.1
HIPPOTraining Paradigm=Ours2026.0661.7
Qwen2.5_7BTraining Paradigm=Standard Training, Configuration=+ RL2026.0661.2
SP3FTraining Paradigm=Reward-Shaping2026.0661.1
CPGPromptBackbone=Qwen3.5-9B2026.0559.15
Original (LLaMA-3.1-8B-Instruct)Backbone=LLaMA-3.1-8B-Instruct2026.0258.7
MedVerseBackbone=Qwen2.5-7B-Instruct2026.0258.6
PRM RLTraining Paradigm=Reward-Shaping2026.0657.7
PREF-GRPOTraining Paradigm=Reward-Shaping2026.0656.6
Original (Qwen2.5-7B-Instruct)Backbone=Qwen2.5-7B-Instruct2026.0256.2
MedReasonBackbone=Qwen2.5-7B-Instruct2026.0256.2
MediPhi-Instruct-4B2026.0553.34
Clinical-R1-3B2026.0552.55
Qwen2.5_7BTraining Paradigm=Standard Training, Configuration=+ SFT2026.0649.5
Qwen2.5_7BTraining Paradigm=Standard Training2026.0649.4
DyLANBase Model=Gemma-3-4B2025.0849.2
ReConcileBase Model=Gemma-3-4B2025.0848.5
Claude-Haiku-4.52026.0548.08
MedAgentsBase Model=Gemma-3-4B2025.0846.8
BioMistral-7B2026.0545.8
TMA-AllComponBase Model=Gemma-3-4B2025.0845.5
Llava-Med-v1.5-7B2026.0544.78
MDAgentsBase Model=Gemma-3-4B2025.0839.6
MedAlpaca-7B2026.0538.81