List your company
ResearchFormerly WizWand

Question Answering on MedMCQA (Accuracy)

80AccuracyClaude Sonnet 4.5
46.7255.366472.64Mar 1, 2026Mar 21, 2026Apr 10, 2026May 1, 2026May 21, 2026Jun 10, 2026Jul 1, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Claude Sonnet 4.5Backbone=Claude Sonnet 4.52026.0780
MedGemma 27BBackbone=MedGemma 27B2026.0773
DeepSeek-R1Backbone=DeepSeek-R12026.0772.8
Qwen3 8BBackbone=Qwen3 8B2026.0766.3
MA-PoP2026.0564.67
TARSEBackbone=14B2026.0364.3
FaithMed (Qwen3 4B)Backbone=Qwen3 4B, Method Variant=FaithMed2026.0764.1
ISP2026.0563.67
Agentic Search (Qwen3 4B)Backbone=Qwen3 4B, Method Variant=Agentic Search2026.0763.4
LLP2026.0563.21
Agentic Search + GRPO (Qwen3 4B)Backbone=Qwen3 4B, Method Variant=Agentic Search + GRPO2026.0763.2
Free MADT=52026.0562.67
TARSEBackbone=7B2026.0362.1
Decentr. MADT=22026.0562
Sparse MADT=22026.0562
Free MADT=32026.0562
AttributionIteration=1, Selection Ratio=0.22026.0461.92
EVOSELECTIteration=2, Selection Ratio=0.52026.0461.87
RandomIteration=1, Selection Ratio=0.22026.0461.8
Attr-DivIteration=1, Selection Ratio=0.52026.0461.8
EVOSELECTIteration=1, Selection Ratio=0.52026.0461.73
Sparse MADT=52026.0561.67
Majority Voting2026.0561.67
Self-Consistency2026.0561.67
Attr-DivIteration=1, Selection Ratio=0.22026.0461.58
AllIteration=22026.0461.58
Attr-DivIteration=2, Selection Ratio=0.52026.0461.58
EVOSELECTIteration=2, Selection Ratio=0.22026.0461.58
DiversityIteration=1, Selection Ratio=0.52026.0461.56
AttributionIteration=2, Selection Ratio=0.52026.0461.53
AttributionIteration=2, Selection Ratio=0.22026.0461.46
AttributionIteration=1, Selection Ratio=0.52026.0461.44
DiversityIteration=2, Selection Ratio=0.52026.0461.44
AllIteration=12026.0461.39
RandomIteration=1, Selection Ratio=0.52026.0461.39
TSDSIteration=1, Selection Ratio=0.52026.0461.39
Free MADT=22026.0561.33
DiversityIteration=1, Selection Ratio=0.22026.0461.32
RandomIteration=2, Selection Ratio=0.22026.0461.32
RandomIteration=2, Selection Ratio=0.52026.0461.32
DiversityIteration=2, Selection Ratio=0.22026.0461.22
Attr-DivIteration=2, Selection Ratio=0.22026.0461.22
EVOSELECTIteration=1, Selection Ratio=0.22026.0461.2
TSDSIteration=2, Selection Ratio=0.52026.0461.13
TSDSIteration=1, Selection Ratio=0.22026.0461.01
Base2026.0460.75
Decentr. MADT=32026.0560.67
TSDSIteration=2, Selection Ratio=0.22026.0460.34
Sparse MADT=32026.0560.33
GoA_MaxAgent Configuration=Multi-Agent, Number of Agents=3, Zero-shot CoT=true2026.0460.04
Decentr. MADT=52026.0560
MedRAG (Qwen3 4B)Backbone=Qwen3 4B, Method Variant=MedRAG2026.0759.8
Base (Qwen3 4B)Backbone=Qwen3 4B, Method Variant=Base2026.0759.7
Centr. MADT=52026.0559.67
Centr. MADT=22026.0559
Gemma 3 12BBackbone=Gemma 3 12B2026.0759
Centr. MADT=32026.0558.33
GoA_MeanAgent Configuration=Multi-Agent, Number of Agents=3, Zero-shot CoT=true2026.0457.92
Falcon-7B2026.0557.67
i-MedRAG-Qwen2.5Backbone=7B2026.0357.6
Llama 3.1 8B InstructBackbone=Llama 3.1 8B2026.0757.3
MedRAG-Qwen2.5Backbone=7B2026.0356.9
FaithMed (Qwen3 1.7B)Backbone=Qwen3 1.7B, Method Variant=FaithMed2026.0756.9
rStar-Qwen2.5Backbone=14B2026.0356.7
rStar-Qwen2.5Backbone=7B2026.0355.9
Qwen2.5-InstructBackbone=14B2026.0355.7
SCAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0455.7
Gemma-9B2026.0555.67
Self-MoAAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0455.56
Qwen2.5-7B-InstructAgent Configuration=Single-Agent, Number of Agents=1, Zero-shot CoT=true, Agent Role/Strategy=General2026.0455.22
RefineAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0454.94
MoAAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0454.94
Qwen-7B2026.0554.67
ReConcileAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0454.6
Qwen2.5-InstructBackbone=7B2026.0354.1
DebateAgent Configuration=Multi-Agent, Number of Agents=6, Zero-shot CoT=true2026.0453.05
MedRAG (Qwen3 1.7B)Backbone=Qwen3 1.7B, Method Variant=MedRAG2026.0751.3
Agentic Search + GRPO (Qwen3 1.7B)Backbone=Qwen3 1.7B, Method Variant=Agentic Search + GRPO2026.0750.1
Llama-8B2026.0550
TSDSIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0449.39
EVOSELECTIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0449.27
EVOSELECTIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0449.15
TSDSIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0448.98
EVOSELECTIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0448.94
RandomIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0448.91
AllIteration=1, Selection Ratio=All, Backbone=3b base model2026.0448.67
EVOSELECTIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0448.63
Attr-DivIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0448.55
DiversityIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0448.48
RandomIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0448.43
Attr-DivIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0448.29
RandomIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0448.24
Attr-DivIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0448.2
AttributionIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0448.17
DiversityIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0448.17
AttributionIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0448.1
AttributionIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0448.08
TSDSIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0448.08
TSDSIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0448.03
Base (Qwen3 1.7B)Backbone=Qwen3 1.7B, Method Variant=Base2026.0748