VLM Robustness on VLM-Bias
20.9AccuracyLLaVA-OV + SFT
Evaluation results
| Method | Links | |
|---|---|---|
| LLaVA-OV + SFTBase Model=LLaVA-OV, Training Stage=SFT2026.07 | 20.9 | |
| LLaVA-OV + IRABase Model=LLaVA-OV, Training Stage=IRA2026.07 | 19.7 | |
| InternVL2 + IRABase Model=InternVL2, Training Stage=IRA2026.07 | 18.4 | |
| InternVL2.5-PtBase Model=InternVL2.5, Training Stage=Pretrain2026.07 | 18.3 | |
| InternVL2.5 + IRABase Model=InternVL2.5, Training Stage=IRA2026.07 | 18.3 | |
| InternVL2 + SFTBase Model=InternVL2, Training Stage=SFT2026.07 | 18.1 | |
| InternVL2.5 + SFTBase Model=InternVL2.5, Training Stage=SFT2026.07 | 17.8 | |
| InternVL2-PtBase Model=InternVL2, Training Stage=Pretrain2026.07 | 17.6 | |
| LLaVA-OV-PtBase Model=LLaVA-OV, Training Stage=Pretrain2026.07 | 16.9 |