Multi-class staging on Dementia Assessment
87.4AUCMacroDementia-Agents
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Dementia-AgentsArchitecture=Multi-Agent2026.06 | 87.4 | 75.7 | 74.5 | |
| LLaVA-Med-7BEvaluation Protocol=Supervised Fine-Tuning, Architecture=Single MLLM2026.06 | 84.1 | 68.9 | 68 | |
| Qwen3-VL-8BEvaluation Protocol=Supervised Fine-Tuning, Architecture=Single MLLM2026.06 | 82.1 | 67 | 65.8 | |
| Hulu-Med-7BEvaluation Protocol=Supervised Fine-Tuning, Architecture=Single MLLM2026.06 | 81.4 | 65 | 63 | |
| InternVL3.5-8BEvaluation Protocol=Supervised Fine-Tuning, Architecture=Single MLLM2026.06 | 78.7 | 64.1 | 59.4 | |
| Qwen3-VL-8BEvaluation Protocol=Few-Shot, Architecture=Single MLLM2026.06 | 72.2 | 60.2 | 60.5 | |
| MedAgentsArchitecture=Multi-Agent2026.06 | 71.8 | 50.5 | 43.1 | |
| InternVL3.5-8BEvaluation Protocol=Zero-Shot, Architecture=Single MLLM2026.06 | 68.9 | 45.6 | 31.6 | |
| Dementia-AgentsArchitecture=Multi-Agent, Evaluation Protocol=Zero-Shot2026.06 | 63.7 | 47.6 | 36.2 | |
| Hulu-Med-7BEvaluation Protocol=Zero-Shot, Architecture=Single MLLM2026.06 | 58.4 | 41.7 | 32.3 | |
| Hulu-Med-7BEvaluation Protocol=Few-Shot, Architecture=Single MLLM2026.06 | 57.8 | 39.8 | 19 | |
| Qwen3-VL-8BEvaluation Protocol=Zero-Shot, Architecture=Single MLLM2026.06 | 52.8 | 41.7 | 26.1 | |
| MedAgentProArchitecture=Multi-Agent2026.06 | 50 | 33 | 16.5 | |
| InternVL3.5-8BEvaluation Protocol=Few-Shot, Architecture=Single MLLM2026.06 | 47.7 | 40.8 | 21.5 | |
| LLaVA-Med-7BEvaluation Protocol=Few-Shot, Architecture=Single MLLM2026.06 | 43.9 | 39.8 | 19 | |
| LLaVA-Med-7BEvaluation Protocol=Zero-Shot, Architecture=Single MLLM2026.06 | 36.9 | 35 | 17.6 |