Grounded Video Question Answering on DeVE-QA 1.0 (test)
0.582mIoPHuman
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Humanmode=Oracle2024.09 | 0.582 | 0.623 | 0.439 | 0.527 | 0.847 | 0.624 | |
| DeVi-Gemini-2.0Supervision=Zero-shot, Backbone=Gemini-2.02024.09 | 0.349 | 0.328 | 0.217 | 0.185 | 0.721 | 0.277 | |
| DeVi-GPT-4oSupervision=Zero-shot, Backbone=GPT-4o2024.09 | 0.338 | 0.322 | 0.207 | 0.174 | 0.719 | 0.271 | |
| LLoViSupervision=Zero-shot2024.09 | 0.275 | 0.27 | 0.179 | 0.13 | 0.639 | 0.229 | |
| SeViLASupervision=Weakly-supervised, Pre-training=video grounding datasets2024.09 | 0.258 | 0.199 | 0.212 | 0.115 | 0.627 | 0.161 | |
| LongVASupervision=Zero-shot2024.09 | 0.249 | 0.247 | 0.169 | 0.152 | 0.662 | 0.208 | |
| Temp[CLIP](NG+)Supervision=Weakly-supervised2024.09 | 0.246 | 0.248 | 0.125 | 0.091 | 0.589 | 0.149 | |
| VideoLLaMA2Supervision=Zero-shot2024.09 | 0.237 | 0.22 | 0.129 | 0.101 | 0.62 | 0.192 | |
| Qwen2-VLSupervision=Zero-shot2024.09 | 0.236 | 0.232 | 0.165 | 0.147 | 0.639 | 0.201 | |
| VideoChat2Supervision=Zero-shot2024.09 | 0.231 | 0.218 | 0.142 | 0.125 | 0.592 | 0.186 | |
| LLaVA-Next-VideoSupervision=Zero-shot2024.09 | 0.225 | 0.211 | 0.138 | 0.107 | 0.569 | 0.174 | |
| FrozenBiLM(NG+)Supervision=Weakly-supervised2024.09 | 0.212 | 0.182 | 0.085 | 0.062 | 0.616 | 0.145 |