Knowledge boundary assessment on V-Idk 3,000 samples (test)
51.9IK-IKDPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DPOModel=LLaVA-1.5 7B, Setting=Zero-shot2026.04 | 51.9 | 5.5 | 57.4 | |
| DPOModel=LLaVA-1.5 13B, Setting=Zero-shot2026.04 | 50.2 | 6.7 | 56.9 | |
| BaseModel=LLaVA-1.5 13B, Setting=Zero-shot2026.04 | 49.9 | 4.3 | 54.2 | |
| DPOModel=LLaVA-1.5 13B, Setting=Few-shot2026.04 | 49.7 | 21.4 | 71.1 | |
| BaseModel=LLaVA-1.5 13B, Setting=Few-shot2026.04 | 49.6 | 19.8 | 69.4 | |
| ORPOModel=LLaVA-1.5 7B, Setting=Zero-shot2026.04 | 45.9 | 14.1 | 60 | |
| BaseModel=LLaVA-1.5 7B, Setting=Zero-shot2026.04 | 45.8 | 1.8 | 47.6 | |
| ORPOModel=LLaVA-1.5 13B, Setting=Zero-shot2026.04 | 45.8 | 20.4 | 66.2 | |
| BaseModel=LLaVA-1.5 7B, Setting=Few-shot2026.04 | 45.5 | 12.4 | 57.9 | |
| DPOModel=LLaVA-1.5 7B, Setting=Few-shot2026.04 | 45.3 | 17.8 | 63.1 | |
| CPOModel=LLaVA-1.5 13B, Setting=Zero-shot2026.04 | 44.9 | 19 | 63.9 | |
| CPOModel=LLaVA-1.5 7B, Setting=Zero-shot2026.04 | 44.4 | 13.9 | 58.3 | |
| CPOModel=LLaVA-1.5 13B, Setting=Few-shot2026.04 | 43.7 | 30.3 | 74 | |
| CPOModel=LLaVA-1.5 7B, Setting=Few-shot2026.04 | 42.8 | 24.5 | 67.3 | |
| ORPOModel=LLaVA-1.5 13B, Setting=Few-shot2026.04 | 42.7 | 33.9 | 76.6 | |
| ORPOModel=LLaVA-1.5 7B, Setting=Few-shot2026.04 | 40 | 26.5 | 66.5 | |
| SFTModel=LLaVA-1.5 7B, Setting=Zero-shot2026.04 | 34.4 | 32.7 | 67.1 | |
| SFTModel=LLaVA-1.5 13B, Setting=Zero-shot2026.04 | 34.4 | 39.1 | 73.5 | |
| SFTModel=LLaVA-1.5 7B, Setting=Few-shot2026.04 | 27.6 | 36.8 | 64.4 | |
| SFTModel=LLaVA-1.5 13B, Setting=Few-shot2026.04 | 27.5 | 39.8 | 67.3 |