Video Question Answering on EgoSchema (official)
76.2AccuracyQwen2.5-VL-72B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-VL-72BMethod Category=Open-Source2025.09 | 76.2 | 180 | |
| CogniGPTMethod Category=Training-Free Agent, Backbone=InternVL2.5-8B2025.09 | 74 | 11.5 | |
| CogniGPTMethod Category=Training-Free Agent, Backbone=Qwen2.5-7B2025.09 | 72.4 | 10.7 | |
| GPT-4oMethod Category=Proprietary2025.09 | 72.2 | 180 | |
| VideoChat-A1Method Category=Open-Source, Backbone=InternVL2.52025.09 | 72.1 | 35.9 | |
| Gemini 1.5-ProMethod Category=Proprietary2025.09 | 71.2 | 180 | |
| VideoChat-A1Method Category=Open-Source, Backbone=Qwen2.5-VL2025.09 | 70.7 | 42 | |
| CogniGPT (Ours)LLM=GPT-42025.09 | 69.2 | 11.2 | |
| DrVideoLLM=GPT-42025.09 | 66.4 | 90 | |
| VideoTreeLLM=GPT-42025.09 | 66.2 | 62.4 | |
| Gemini 1.5-FlashMethod Category=Proprietary2025.09 | 65.7 | 180 | |
| Qwen2.5-VL-7BMethod Category=Open-Source2025.09 | 65 | 180 | |
| VideoAgentLLM=GPT-42025.09 | 62.8 | 90 | |
| LLaVA-OneVision-72BMethod Category=Open-Source2025.09 | 62 | 32 | |
| VideoAgent*LLM=GPT-42025.09 | 60.2 | 8.4 | |
| LLoViLLM=GPT-3.52025.09 | 51.8 | 180 |