Hallucination Evaluation on FINER-DOCCI 3K MCQs per setting 1.0
65.9Multi-Object Paired AccuracyInternVL-3.5 + FINER-Tuning
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InternVL-3.5 + FINER-TuningSize=14B, tuning=FINER-Tuning2026.03 | 65.9 | 65 | 57 | 23 | |
| Gemini-2.5-FlashSize=-2026.03 | 64.4 | 64.5 | 56.7 | 49.6 | |
| InternVL-3.5 + FINER-TuningSize=8B, tuning=FINER-Tuning2026.03 | 62.6 | 60.1 | 52.7 | 23.7 | |
| InternVL-3.5Size=38B2026.03 | 62.3 | 64.8 | 54.2 | 36.6 | |
| InternVL-3.5Size=14B2026.03 | 58.6 | 55.9 | 41.4 | 15.6 | |
| InternVL-3.5Size=8B2026.03 | 58.1 | 54.3 | 41.8 | 16.8 | |
| Qwen2.5-VL + FINER-TuningSize=7B, tuning=FINER-Tuning2026.03 | 49.8 | 52.2 | 43.4 | 28 | |
| Qwen2.5-VLSize=7B2026.03 | 48.7 | 47.5 | 36.7 | 23.4 | |
| RLAIF-VSize=12B2026.03 | 46.5 | 31.7 | 32.4 | 19.4 | |
| LLaVA-1.6 + FINER-TuningSize=7B, tuning=FINER-Tuning2026.03 | 26.4 | 29.4 | 24.7 | 18.5 | |
| RLHF-VSize=13B2026.03 | 13.2 | 7.2 | 8.1 | 7 | |
| LLaVA-1.6Size=7B2026.03 | 10.1 | 12.3 | 8.2 | 13.3 | |
| OPA-DPOSize=13B2026.03 | 8.1 | 5.5 | 8.3 | 8 | |
| LLaVA-RLHFSize=13B2026.03 | 7.3 | 3 | 5.1 | 5.3 | |
| LRV-V2Size=13B2026.03 | 6.3 | 5.4 | 6.1 | 5.2 | |
| Random GuessSize=-2026.03 | 4 | 4 | 4 | 4 |