Video Question Answering on Video-MME Overall 1.0
59Accuracy (No Subtitles)VILA-1.5
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VILA-1.5LLM Size=34B2024.11 | 59 | 59.4 | |
| LLaVA-NeXT-VideoLLM Params=34B2024.11 | 52 | 54.9 | |
| LLaVA-NeXTLLM Size=34B, DPO tuning=true2024.11 | 52 | 54.9 | |
| Qwen-VL-Max2024.11 | 51.3 | 51.2 | |
| InternVL-V1.5LLM Params=20B2024.11 | 50.7 | 52.4 | |
| PPLLAVALLM Params=7B, Multi-image training=true2024.11 | 48.8 | 53.6 | |
| PPLLAVALLM Params=7B, Multi-image training=false2024.11 | 46.1 | 50 | |
| DynFocusInput=16 f, LLM Size=7B, Resolution=224x224, L=25, K/L=0.82024.11 | 44.1 | 47.8 | |
| Qwen-VL-ChatLLM Params=7B2024.11 | 41.4 | 41.9 | |
| LLaVA-NeXT-VideoLLM Params=7B2024.11 | 40.9 | 45 | |
| Chat-UniVi-V1.5LLM Params=7B2024.11 | 40.6 | 45.9 | |
| Chat-UniViLLM Size=7B2024.11 | 40.6 | 45.9 | |
| Video-LLaVALLM Params=7B2024.11 | 39.9 | 41.6 | |
| Video-LLaVAInput=8 f, LLM Size=7B2024.11 | 39.9 | 41.6 | |
| VideoChat2-MistralLLM Params=7B2024.11 | 39.5 | 43.8 | |
| VideoChat2Input=16 f, LLM Size=7B2024.11 | 39.5 | 43.8 | |
| ST-LLMLLM Params=7B2024.11 | 37.9 | 42.3 | |
| ST-LLMInput=16 f, LLM Size=7B2024.11 | 37.9 | 42.3 | |
| LLaMA-VIDInput=1 fps, LLM Size=7B2024.11 | 25.9 | — |