Conflict Discrimination on REAL-VQA (test)
98.1MCCRPA-SFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RPA-SFTbackbone=Qwen3-VL-8B2026.02 | 98.1 | 99.1 | |
| RPA-SFTbackbone=InternVL3.5-8B2026.02 | 95.6 | 98.5 | |
| RPA-SFTbackbone=LLaVA-1.5-7B2026.02 | 89.4 | 89.6 | |
| SFTbackbone=Qwen3-VL-8B2026.02 | 89.4 | 96.8 | |
| SFTbackbone=InternVL3.5-8B2026.02 | 88.4 | 96.1 | |
| SFTbackbone=LLaVA-1.5-7B2026.02 | 68.1 | 84.6 | |
| Few-shot CoTbackbone=Qwen3-VL-8B2026.02 | 19.4 | 72.3 | |
| Zero-shotbackbone=Qwen3-VL-8B2026.02 | 19 | 69.9 | |
| Few-shot CoTbackbone=InternVL3.5-8B2026.02 | 11.5 | 70.8 | |
| Zero-shotbackbone=InternVL3.5-8B2026.02 | 9.3 | 70.6 | |
| Zero-shotbackbone=LLaVA-1.5-7B2026.02 | 2.9 | 52.3 | |
| Few-shot CoTbackbone=LLaVA-1.5-7B2026.02 | -3.7 | 47.4 |