Long-form Video Question Answering on OPENTSUBENCH
72.1Object Centric AccuracyQwen2.5-VL
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen2.5-VLLLM=Qwen 2.52026.06 | 72.1 | 40.9 | 26.6 | 27.2 | 35.8 | 39.3 | |
| TimeSuiteLLM=Mistral-7B2026.06 | 68.2 | 31.4 | 23.9 | 80.9 | 26.7 | 35.4 | |
| VideoagentLLM=GPT-4o, VLM=GPT-4o2026.06 | 65.1 | 35.6 | 17.5 | 34.9 | 28.6 | 33.9 | |
| VideoChat-FlashLLM=InternLM22026.06 | 63.6 | 36.4 | 29.3 | 66.7 | 29.7 | 37.8 | |
| TimeChatLLM=LLaMA-2-7B2026.06 | 62.8 | 30.7 | 13.1 | 55.7 | 21.2 | 30.4 | |
| VTimeLLMLLM=LLaMA-2-7B2026.06 | 55.8 | 27 | 32.5 | 65.9 | 29.7 | 33.1 | |
| TIMEPROVELLM=Gemma4-2B, VLM=VLMA32026.06 | 53.9 | 33.5 | 27.7 | 82.9 | 31.7 | 37.3 | |
| TIMEPROVELLM=Qwen2-7B, VLM=VLMA32026.06 | 53.5 | 42 | 31.7 | 78.9 | 36.1 | 42.7 | |
| Time-R1LLM=Qwen2.5VL2026.06 | 50.8 | 35.6 | 26.9 | 49.2 | 28.8 | 34.9 | |
| TIMEPROVELLM=Gemma4-2B, VLM=GPT-4o2026.06 | 49.2 | 47.9 | 37.1 | 70.3 | 35.7 | 45.1 | |
| VideoTreeLLM=GPT-5.12026.06 | 36.8 | 30.3 | 19.4 | 33.7 | 21.2 | 27.3 | |
| GPT-4oLLM=GPT-4o, VLM=GPT-4o2026.06 | 27.9 | 27.3 | 9.4 | 67.7 | 11.2 | 23.8 | |
| AVPLLM=GPT-4o, VLM=GPT-4o2026.06 | 11.5 | 20.2 | 9.4 | 42.3 | 3.9 | 14.4 | |
| VideoLLaMA3LLM=Qwen 22026.06 | 7.8 | 22.3 | 21.5 | 71.9 | 15.7 | 22.6 |