Video Question Answering on VidHal
0.863Object AccuracySANTA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SANTAbase_model=LLaVA-Video2025.12 | 0.863 | 0.858 | 0.861 | |
| LLaVA-Video2025.12 | 0.843 | 0.831 | 0.837 | |
| LLaVA-Video + DeCoadaptation=DeCo2025.12 | 0.828 | 0.847 | 0.838 | |
| LLaVA-Video + HALVAadaptation=HALVA2025.12 | 0.799 | 0.809 | 0.804 | |
| LLaVA-Video + HACLadaptation=HACL2025.12 | 0.799 | 0.842 | 0.821 | |
| LLaVA-Video + SFTadaptation=SFT2025.12 | 0.363 | 0.656 | 0.51 |