Long Video Question Answering on TVQA Long
52.1Overall AccuracyLLaVA-Video + OneClip-RAG
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-Video + OneClip-RAGLLM=7B, # Frames=642025.12 | 52.1 | |
| AKSLLM=7B, # Frames=642025.12 | 49.9 | |
| LLaVA-VideoLLM=7B, # Frames=642025.12 | 49.6 | |
| Video-XLLLM=7B, # Frames=1282025.12 | 43.9 | |
| LongVALLM=7B, # Frames=2562025.12 | 43.3 | |
| Qwen2.5-VLLLM=7B, # Frames=322025.12 | 42.3 |