List your company

Medical Reasoning on MedXpertQA

42.9AccuracyDeepSeek-R1
7.95617.02826.135.172Sep 28, 2025Nov 13, 2025Dec 29, 2025Feb 13, 2026Mar 31, 2026May 16, 2026Jul 1, 2026

Evaluation results

MethodLinks
DeepSeek-R1Backbone=DeepSeek-R12026.0742.9
Claude Sonnet 4.5Backbone=Claude Sonnet 4.52026.0739.4
MedLABackbone Model=deepseek-r1, Sample Count=602025.0936
FaithMed (Qwen3 4B)Backbone=Qwen3 4B2026.0728.2
MedLABackbone Model=deepseek-v3, Sample Count=602025.0925.6
Agentic Search + GRPO (Qwen3 4B)Backbone=Qwen3 4B2026.0722
baselineBackbone Model=deepseek-r1, Sample Count=602025.0921.3
Agentic Search (Qwen3 4B)Backbone=Qwen3 4B2026.0719.6
Qwen3 8BBackbone=Qwen3 8B2026.0719.2
MedGemma 27BBackbone=MedGemma 27B2026.0719
MedRAG (Qwen3 4B)Backbone=Qwen3 4B2026.0719
FaithMed (Qwen3 1.7B)Backbone=Qwen3 1.7B2026.0717.1
Gemma 3 12BBackbone=Gemma 3 12B2026.0716.4
Base (Qwen3 4B)Backbone=Qwen3 4B2026.0716
baselineBackbone Model=deepseek-v3, Sample Count=602025.0915
MedRAG (Qwen3 1.7B)Backbone=Qwen3 1.7B2026.0714.2
Llama 3.1 8B InstructBackbone=Llama 3.1 8B2026.0711.8
Agentic Search + GRPO (Qwen3 1.7B)Backbone=Qwen3 1.7B2026.0711.5
Agentic Search (Qwen3 1.7B)Backbone=Qwen3 1.7B2026.0711.1
Base (Qwen3 1.7B)Backbone=Qwen3 1.7B2026.079.3