Multimodal Medical Reasoning on VQA-RAD
80.45Accuracy (%)Dialectic-Med
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Dialectic-MedBackbone=GPT-5.12026.04 | 80.45 | 4,500 | 22.5 | |
| Dialectic-MedBackbone=Gemini-3 Pro2026.04 | 79.6 | 4,000 | 16 | |
| VISTA-SFTBackbone=InternVL3-8B2026.05 | 78.49 | — | — | |
| Dialectic-MedBackbone=GPT-4o2026.04 | 78.24 | 3,800 | 15.2 | |
| Dialectic-MedBackbone=Claude-4.5 Sonnet2026.04 | 76.65 | 4,200 | 22.68 | |
| SFT-OracleBackbone=InternVL3-8B2026.05 | 76.49 | — | — | |
| GPT-5.1Category=CoT Baseline2026.04 | 76.4 | 1,200 | 6 | |
| SFT-SeedBackbone=InternVL3-8B2026.05 | 75.7 | — | — | |
| Gemini-3 ProCategory=CoT Baseline2026.04 | 75.15 | 1,100 | 4.4 | |
| MedMMVCategory=Medical Agents2026.04 | 74.87 | 9,000 | 36 | |
| Dialectic-MedBackbone=Qwen3-VL-32B2026.04 | 74.62 | 3,000 | 0.35 | |
| STaRBackbone=InternVL3-8B2026.05 | 74.5 | — | — | |
| GPT-4oCategory=CoT Baseline2026.04 | 74.28 | 1,000 | 4 | |
| MedVCDCategory=Medical Agents2026.04 | 73.2 | 8,500 | 34 | |
| Claude-4.5 SonnetCategory=CoT Baseline2026.04 | 72.82 | 1,200 | 6.48 | |
| VISTA-SFTBackbone=InternVL3-2B2026.05 | 72.51 | — | — | |
| STaR+Backbone=InternVL3-8B2026.05 | 72.51 | — | — | |
| R3VBackbone=InternVL3-8B2026.05 | 72.51 | — | — | |
| DS-RL + GTPModel=Lingshu, Scale=32B2026.06 | 72.51 | — | — | |
| R3VBackbone=InternVL3-2B2026.05 | 72.31 | — | — | |
| SFT-OracleBackbone=InternVL3-2B2026.05 | 72.11 | — | — | |
| GTPModel=Lingshu, Scale=7B2026.06 | 71.71 | — | — | |
| ReConcileCategory=Medical Agents2026.04 | 71.31 | 12,000 | 48 | |
| VISTA-SFTBackbone=Qwen3-VL-2B-Instruct2026.05 | 71.31 | — | — | |
| DS-RL + GTPModel=Lingshu, Scale=7B2026.06 | 71.31 | — | — | |
| DS-RL + GTPModel=HuatuoGPT-Vision, Scale=7B2026.06 | 70.52 | — | — | |
| Instruct ModelModel=Lingshu, Scale=32B2026.06 | 70.52 | — | — | |
| DS-RLModel=Lingshu, Scale=32B2026.06 | 70.52 | — | — | |
| Dialectic-MedBackbone=Qwen3-VL-8B2026.04 | 70.35 | 2,500 | 0.15 | |
| STaR+Backbone=Qwen3-VL-2B-Instruct2026.05 | 69.32 | — | — | |
| DS-RLModel=Lingshu, Scale=7B2026.06 | 69.32 | — | — | |
| GTPModel=Lingshu, Scale=32B2026.06 | 69.32 | — | — | |
| Qwen3-VL-32BCategory=CoT Baseline2026.04 | 68.96 | 900 | 0.1 | |
| SFT-SeedBackbone=Qwen3-VL-2B-Instruct2026.05 | 68.92 | — | — | |
| SFT-OracleBackbone=Qwen3-VL-2B-Instruct2026.05 | 68.92 | — | — | |
| STaR+Backbone=InternVL3-2B2026.05 | 68.92 | — | — | |
| Med-PaLM 2Category=CoT Baseline2026.04 | 68.65 | 800 | 1.5 | |
| DS-RLModel=HuatuoGPT-Vision, Scale=7B2026.06 | 68.13 | — | — | |
| STaRBackbone=InternVL3-2B2026.05 | 66.53 | — | — | |
| Instruct ModelModel=HuatuoGPT-Vision, Scale=7B2026.06 | 66.53 | — | — | |
| R3VBackbone=Qwen3-VL-2B-Instruct2026.05 | 66.14 | — | — | |
| STaRBackbone=Qwen3-VL-2B-Instruct2026.05 | 65.34 | — | — | |
| GTPModel=HuatuoGPT-Vision, Scale=7B2026.06 | 65.34 | — | — | |
| MedAgentCategory=Medical Agents2026.04 | 65.33 | 15,000 | 60 | |
| Instruct ModelModel=Lingshu, Scale=7B2026.06 | 64.14 | — | — | |
| Qwen3-VL-8BCategory=CoT Baseline2026.04 | 62.5 | 800 | 0.05 | |
| SFT-SeedBackbone=InternVL3-2B2026.05 | 58.96 | — | — | |
| LLaVA-MedCategory=CoT Baseline2026.04 | 56.2 | 700 | 0.05 |