Multimodal Evaluation on SeedBench (SeedBench Metric)
78SeedBench ScoreSFT + VIGIL (Ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT + VIGIL (Ours)Base Model=InternVL2.5-26B [4]2026.06 | 78 | |
| SFT + DA-DPOBase Model=InternVL2.5-26B [4]2026.06 | 77.2 | |
| SFT + DPOBase Model=InternVL2.5-26B [4]2026.06 | 76.8 | |
| SFT onlyBase Model=InternVL2.5-26B [4]2026.06 | 76.4 | |
| SFT + VIGIL (Ours)Base Model=LLaVA-OneVision-7B [3]2026.06 | 73.1 | |
| SFT + DA-DPOBase Model=LLaVA-OneVision-7B [3]2026.06 | 72.4 | |
| SFT + DPOBase Model=LLaVA-OneVision-7B [3]2026.06 | 71.8 | |
| SFT onlyBase Model=LLaVA-OneVision-7B [3]2026.06 | 71.5 |