List your company

Medical Visual Question Answering / Tool-use on ChestAgentBench OOD (test)

76.2AccuracyGemini-3-flash
45.41653.40861.469.392Mar 9, 2026

Evaluation results

MethodLinks
Gemini-3-flashFramework=Framework I, Inference Mode=Direct Inference2026.0376.2
Gemini-3-flashFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.0372.7
GPT-4oFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.0363.1
MeissaFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.0362.8
GPT-4oFramework=Framework I, Inference Mode=Direct Inference2026.0356.4
Qwen3-VL-4BFramework=Framework I, Inference Mode=Direct Inference, Training Protocol=SFT2026.0353.3
Qwen3-VL-4BFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.0346.6