Multimodal Evaluation on MMMU
57.6ScoreVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaBackbone=InternVL3.5-8B2026.05 | 57.6 | |
| EVABackbone=InternVL3.5-8B2026.05 | 57.6 | |
| VanillaBackbone=InternVL3.5-8B2026.05 | 57.6 | |
| EVABackbone=InternVL3.5-8B2026.05 | 57.6 | |
| VLGuardBackbone=InternVL3.5-8B2026.05 | 57.1 | |
| VLGuardBackbone=InternVL3.5-8B2026.05 | 57.1 | |
| EVA(text)Backbone=InternVL3.5-8B2026.05 | 57 | |
| EVA(text)Backbone=InternVL3.5-8B2026.05 | 57 | |
| JailGuardBackbone=InternVL3.5-8B2026.05 | 55.8 | |
| JailGuardBackbone=InternVL3.5-8B2026.05 | 55.8 | |
| VLGuardBackbone=Qwen2.5-VL-7B2026.05 | 50.2 | |
| VLGuardBackbone=Qwen2.5-VL-7B2026.05 | 50.2 | |
| VanillaBackbone=Qwen2.5-VL-7B2026.05 | 49.9 | |
| VanillaBackbone=Qwen2.5-VL-7B2026.05 | 49.9 | |
| EVA(text)Backbone=Qwen2.5-VL-7B2026.05 | 49.6 | |
| EVA(text)Backbone=Qwen2.5-VL-7B2026.05 | 49.6 | |
| EVABackbone=Qwen2.5-VL-7B2026.05 | 48.2 | |
| EVABackbone=Qwen2.5-VL-7B2026.05 | 48.2 | |
| JailGuardBackbone=Qwen2.5-VL-7B2026.05 | 47.5 | |
| JailGuardBackbone=Qwen2.5-VL-7B2026.05 | 47.5 | |
| Adashield-ABackbone=InternVL3.5-8B2026.05 | 38.4 | |
| Adashield-ABackbone=InternVL3.5-8B2026.05 | 38.4 | |
| Adashield-ABackbone=Qwen2.5-VL-7B2026.05 | 35.4 | |
| Adashield-ABackbone=Qwen2.5-VL-7B2026.05 | 35.4 | |
| EVABackbone=LLaVA-1.5-7B2026.05 | 33.4 | |
| EVABackbone=LLaVA-1.5-7B2026.05 | 33.4 | |
| EVA(text)Backbone=LLaVA-1.5-7B2026.05 | 33.3 | |
| EVA(text)Backbone=LLaVA-1.5-7B2026.05 | 33.3 | |
| VLGuardBackbone=LLaVA-1.5-7B2026.05 | 32.1 | |
| VLGuardBackbone=LLaVA-1.5-7B2026.05 | 32.1 | |
| VanillaBackbone=LLaVA-1.5-7B2026.05 | 31.3 | |
| VanillaBackbone=LLaVA-1.5-7B2026.05 | 31.3 | |
| JailGuardBackbone=LLaVA-1.5-7B2026.05 | 29.8 | |
| JailGuardBackbone=LLaVA-1.5-7B2026.05 | 29.8 | |
| Adashield-ABackbone=LLaVA-1.5-7B2026.05 | 19.7 | |
| Adashield-ABackbone=LLaVA-1.5-7B2026.05 | 19.7 |