Vision-Language Medical Reasoning on PathVQA
85.37Accuracy (%)DS-RL + GTP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DS-RL + GTPModel=Lingshu, Scale=32B2026.06 | 85.37 | — | — | |
| DS-RLModel=Lingshu, Scale=32B2026.06 | 85.34 | — | — | |
| GTPModel=Lingshu, Scale=32B2026.06 | 85.19 | — | — | |
| Instruct ModelModel=Lingshu, Scale=32B2026.06 | 77.28 | — | — | |
| DS-RL + GTPModel=Lingshu, Scale=7B2026.06 | 77.19 | — | — | |
| DS-RLModel=Lingshu, Scale=7B2026.06 | 77.16 | — | — | |
| GTPModel=Lingshu, Scale=7B2026.06 | 74.93 | — | — | |
| Dialectic-MedBackbone=GPT-5.12026.04 | 72.32 | 4.5 | 22.5 | |
| Instruct ModelModel=Lingshu, Scale=7B2026.06 | 71.98 | — | — | |
| Dialectic-MedBackbone=Gemini-3 Pro2026.04 | 71.45 | 4 | 16 | |
| Dialectic-MedBackbone=GPT-4o2026.04 | 70.08 | 3.8 | 15.2 | |
| Dialectic-MedBackbone=Claude-4.5 Sonnet2026.04 | 69.92 | 4.2 | 22.68 | |
| GPT-5.1Category=CoT Baseline2026.04 | 68.92 | 1.2 | 6 | |
| MedMMVCategory=Medical Agents2026.04 | 68.17 | 9 | 36 | |
| Dialectic-MedBackbone=Qwen3-VL-32B2026.04 | 67.4 | 3 | 0.35 | |
| Gemini-3 ProCategory=CoT Baseline2026.04 | 67.3 | 1.1 | 4.4 | |
| MedVCDCategory=Medical Agents2026.04 | 67.06 | 8.5 | 34 | |
| GPT-4oCategory=CoT Baseline2026.04 | 66.62 | 1 | 4 | |
| Instruct ModelModel=HuatuoGPT-Vision, Scale=7B2026.06 | 65.76 | — | — | |
| Claude-4.5 SonnetCategory=CoT Baseline2026.04 | 65.64 | 1.2 | 6.48 | |
| ReConcileCategory=Medical Agents2026.04 | 65.12 | 12 | 48 | |
| DS-RLModel=HuatuoGPT-Vision, Scale=7B2026.06 | 64.99 | — | — | |
| Dialectic-MedBackbone=Qwen3-VL-8B2026.04 | 64.15 | 2.5 | 0.15 | |
| DS-RL + GTPModel=HuatuoGPT-Vision, Scale=7B2026.06 | 63.8 | — | — | |
| Qwen3-VL-32BCategory=CoT Baseline2026.04 | 62.33 | 0.9 | 0.1 | |
| MedAgentCategory=Medical Agents2026.04 | 62.17 | 15 | 60 | |
| GTPModel=HuatuoGPT-Vision, Scale=7B2026.06 | 62.05 | — | — | |
| Med-PaLM 2Category=CoT Baseline2026.04 | 60.08 | 0.8 | 1.5 | |
| Qwen3-VL-8BCategory=CoT Baseline2026.04 | 58.31 | 0.8 | 0.05 | |
| LLaVA-MedCategory=CoT Baseline2026.04 | 54.45 | 0.7 | 0.05 | |
| DyLANModel=Gemma-3-4B, Framework=DyLAN2025.08 | — | 7,901 | — | |
| DyLANModel=GPT-4o, Framework=DyLAN2025.08 | — | 4,301 | — | |
| MDAgentsModel=Gemma-3-4B, Framework=MDAgents2025.08 | — | 17,927 | — | |
| MDAgentsModel=GPT-4o, Framework=MDAgents2025.08 | — | 17,927 | — | |
| MedAgentsModel=Gemma-3-4B, Framework=MedAgents2025.08 | — | 23,226 | — | |
| MedAgentsModel=GPT-4o, Framework=MedAgents2025.08 | — | 23,226 | — | |
| ReConcileModel=Gemma-3-4B, Framework=ReConcile2025.08 | — | 3,253 | — | |
| ReConcileModel=GPT-4o, Framework=ReConcile2025.08 | — | 7,325 | — | |
| TMA-AllComponModel=MedGemma-4B, Framework=TMA-AllCompon2025.08 | — | 3,648 | — | |
| TMA-AllComponModel=Gemma-3-4B, Framework=TMA-AllCompon2025.08 | — | 3,648 | — | |
| TMA-AllComponModel=GPT-4o, Framework=TMA-AllCompon2025.08 | — | 3,648 | — |