Pathological Multimodal Understanding on PathMMU (test)
70.9ACSGPT-5
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-5Framework=OpenAI-Agents-SDK, Planner=GPT-52026.02 | 70.9 | 75.5 | 77.3 | 48 | 71.9 | 28.8 | 0 | |
| LAMMIFramework=LAMMI (Ours), Planner=Qwen3 VL 8B Instruct, Adaptation=TA+PE2026.02 | 58.2 | 54.6 | 56.1 | 38.4 | 52.1 | 45.5 | 1.7 | |
| Qwen3 VL 8B InstructFramework=ReACT, Planner=Qwen3 VL 8B Instruct2026.02 | 57.2 | 43.3 | 54.1 | 28.8 | 51.8 | 48.3 | 0.1 | |
| Qwen3 8B + ImageBindFramework=MLLM-Tools, Planner=Qwen3 8B + ImageBind2026.02 | 57.1 | 27.7 | 31 | 20.6 | 27.5 | 57.9 | 75 | |
| MiniCPM-V 4.5Framework=ReACT, Planner=MiniCPM-V 4.52026.02 | 55.4 | 43.8 | 49.4 | 23.2 | 51.2 | 53.8 | 9.2 | |
| LAMMIFramework=LAMMI (Ours), Planner=MiniCPM-V 4.5, Adaptation=TA+PE2026.02 | 54.1 | 49 | 46.9 | 29.9 | 49.8 | 28.1 | 13.9 | |
| LAMMIFramework=LAMMI (Ours), Planner=InternVL3.5 8B, Adaptation=TA+PE2026.02 | 52.8 | 38.2 | 39 | 18.1 | 51.2 | 67.1 | 0 | |
| InternVL3.5 8BFramework=MAT-Agent, Planner=InternVL3.5 8B2026.02 | 50.6 | 34.9 | 43.7 | 25.7 | 44 | 72.6 | 4.1 | |
| InternVL3.5 8BFramework=ReACT, Planner=InternVL3.5 8B2026.02 | 50 | 34.1 | 36.9 | 19.8 | 44.3 | 69.1 | 0 | |
| Qwen3 VL 8B InstructFramework=MAT-Agent, Planner=Qwen3 VL 8B Instruct2026.02 | 46.9 | 32.2 | 37.6 | 22.6 | 39.5 | 69.4 | 12.8 | |
| Vicuna v1.5 7B + ImageBindFramework=MLLM-Tools, Planner=Vicuna v1.5 7B + ImageBind2026.02 | 42.6 | 29.8 | 29.2 | 18.6 | 27.5 | 72.7 | 4.5 | |
| ToolACE 8B + ImageBindFramework=MLLM-Tools, Planner=ToolACE 8B + ImageBind2026.02 | 40.5 | 25.6 | 25.9 | 21 | 24.9 | 77.8 | 41.3 | |
| MiniCPM-V 4.5Framework=MAT-Agent, Planner=MiniCPM-V 4.52026.02 | 40.3 | 23.1 | 26.1 | 14.1 | 31.5 | 74.2 | 17.3 |