Multimodal Reasoning Across University-Level Disciplines on MMMU
54.2AccuracyMMS-PRM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MMS-PRMSize=8B2026.06 | 54.2 | 73 | |
| InternVL-2.5Size=8B2026.06 | 54.1 | 68 | |
| InternVL2.5-MPOSize=8B2026.06 | 53.8 | 69.4 | |
| Qwen2-VLSize=7B2026.06 | 53.7 | 65.1 | |
| InternVL2.5-MPO + SFTSize=8B2026.06 | 53.7 | 71 | |
| Insight-VSize=7B2026.06 | 50.2 | 65.7 | |
| MiniCPM-V-2-6Size=8B2026.06 | 49.8 | 64.2 | |
| LLaVA-OneVisionSize=7B2026.06 | 48.8 | 64.6 | |
| Ovis1.5-LLaMA3Size=8B2026.06 | 48.3 | — | |
| IXC-2.5Size=7B2026.06 | 42.9 | — | |
| LLaVA-NeXT-LLaMA3 + SFTSize=8B2026.06 | 39.6 | 60.8 | |
| LLaVA-NeXT-LLaMA3Size=8B2026.06 | 36.9 | 52.1 |