Grounded Question Answering on NExT-GQA (test)
27.6Accuracy @ GQAVITED
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| VITEDLLM=LLaMA-22025.03 | 27.6 | — | — | — | 41.2 | — | — | — | |
| LLoViSupervision Type=Zero-shot, LM=GPT-4, Params=1.8T2023.12 | 26.8 | — | 39.4 | — | 38 | 21.5 | — | 16.2 | |
| LLoViLLM=GPT-42025.03 | 26.8 | — | — | — | 38 | — | — | — | |
| VideoStreamingParams=7B+1.3B, Specialized grounding module=false2024.05 | 17.8 | — | 32.2 | — | 31 | 19.3 | 13.3 | — | |
| FrozenBiLMSupervision Type=Weakly-Supervised, LM=DEBERTa, Params=900M2023.12 | 17.5 | — | 24.2 | — | 23.7 | 9.6 | — | 6.1 | |
| FrozenBiLMLLM=DeBERTa2025.03 | 17.5 | — | — | — | 23.7 | — | — | — | |
| LangRepoParams=8×7B, Specialized grounding module=false2024.05 | 17.1 | — | 31.3 | — | 28.7 | 18.5 | 12.2 | — | |
| SeViLAParams=4B, Specialized grounding module=true2024.05 | 16.6 | — | 29.5 | — | 22.9 | 21.7 | 13.8 | — | |
| SeViLASupervision Type=Weakly-Supervised, LM=Flan-T5, Params=4B2023.12 | 16.6 | — | 29.5 | — | 22.9 | 21.7 | — | 13.8 | |
| SeViLALLM=Flan-T52025.03 | 16.6 | — | — | — | 22.9 | — | — | — | |
| TempCLIPParams=130M, Specialized grounding module=true2024.05 | 16 | — | 25.7 | — | 25.5 | 12.1 | 8.9 | — | |
| Temp[CLIP]Supervision Type=Weakly-Supervised, LM=Transformer, Params=130M2023.12 | 16 | — | 25.7 | — | 25.5 | 12.1 | — | 8.9 | |
| LLaMA-3.2VLLM=LLaMA-32025.03 | 11.6 | — | — | — | 19.8 | — | — | — | |
| LLoViParams=7B, Specialized grounding module=false2024.05 | 11.2 | — | 20.7 | — | 20.5 | 8.7 | 6 | — | |
| LangRepoParams=7B, Specialized grounding module=false2024.05 | 11.2 | — | 20.3 | — | 20 | 8.7 | 6 | — | |
| IGVSupervision Type=Weakly-Supervised, Params=110M2023.12 | 10.2 | — | 21.4 | — | 18.9 | 14 | — | 9.6 | |
| VideoChat-TPO2024.12 | 0.255 | 0.777 | 0.356 | 0.475 | 0.328 | — | — | — | |
| LLoVi2024.12 | 0.243 | 0.659 | 0.373 | — | 0.369 | — | — | — | |
| VideoStreaming2024.12 | 0.178 | 0.574 | 0.322 | — | 0.31 | — | — | — | |
| FrozenBiLM NG+2024.12 | 0.175 | 0.738 | 0.242 | 0.285 | 0.237 | — | — | — | |
| LangRepo2024.12 | 0.171 | 0.596 | 0.313 | — | 0.287 | — | — | — | |
| SeViLA2024.12 | 0.166 | 0.725 | 0.295 | 0.347 | 0.229 | — | — | — | |
| VIOLETv22024.12 | 0.128 | 0.549 | 0.236 | 0.251 | 0.233 | — | — | — |