Advertising Video Understanding on AdsQA
0.328Strict Accuracy - VUAD-MIR
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AD-MIRModel Type=Reasoning Agent/Model, Backbone=o12026.02 | 0.328 | 0.361 | 0.416 | 0.379 | 0.381 | 0.381 | 0.56 | 0.599 | 0.632 | 0.588 | 0.606 | 0.6 | |
| o1Model Type=Commercial Large Multimodal Model2026.02 | 0.318 | 0.32 | 0.349 | 0.31 | 0.32 | 0.331 | 0.509 | 0.528 | 0.548 | 0.5 | 0.512 | 0.524 | |
| AD-MIRModel Type=Reasoning Agent/Model, Backbone=Qwen2.5-VL-7B2026.02 | 0.316 | 0.34 | 0.31 | 0.295 | 0.274 | 0.307 | 0.525 | 0.559 | 0.53 | 0.502 | 0.501 | 0.521 | |
| DVDModel Type=Reasoning Agent/Model, Backbone=o12026.02 | 0.311 | 0.34 | 0.398 | 0.359 | 0.366 | 0.363 | 0.45 | 0.509 | 0.548 | 0.485 | 0.515 | 0.505 | |
| Qwen2.5-VL-7BModel Type=Open-sourced Video-LLM, Backbone=7B2026.02 | 0.276 | 0.271 | 0.257 | 0.263 | 0.25 | 0.262 | 0.498 | 0.516 | 0.488 | 0.485 | 0.472 | 0.49 | |
| ReAd-RModel Type=Reasoning Agent/Model, Backbone=Qwen2.5-VL-7B2026.02 | 0.204 | 0.279 | 0.272 | 0.221 | 0.255 | 0.25 | 0.462 | 0.562 | 0.546 | 0.482 | 0.526 | 0.515 | |
| LLaVA-Video-7BModel Type=Open-sourced Video-LLM, Backbone=7B2026.02 | 0.14 | 0.144 | 0.187 | 0.152 | 0.173 | 0.161 | 0.378 | 0.416 | 0.451 | 0.381 | 0.436 | 0.41 | |
| LLaVA-OneVision-7BModel Type=Open-sourced Video-LLM, Backbone=7B2026.02 | 0.118 | 0.116 | 0.162 | 0.137 | 0.151 | 0.14 | 0.358 | 0.395 | 0.432 | 0.368 | 0.417 | 0.391 | |
| VideoLLaMA2-7BModel Type=Open-sourced Video-LLM, Backbone=7B2026.02 | 0.0456 | 0.0775 | 0.0728 | 0.0606 | 0.0738 | 0.0648 | 0.216 | 0.29 | 0.284 | 0.226 | 0.272 | 0.252 |