Temporal Grounding on Charades-CON
83.3Ground ScoreTimeChat
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| TimeChatData=VTune, FT=TCAS2025.10 | 83.3 | 75 | — | 39.5 | — | 52.9 | — | 50.8 | — | |
| TimeChatData=VTune, FT=SFT2025.10 | 76.2 | 69.2 | — | 36.2 | — | 44.8 | — | 42.4 | — | |
| TimeChatData=TimeIT, FT=TCAS2025.10 | 66.2 | 60.7 | — | 33.8 | — | 37.3 | — | 31.4 | — | |
| videollamaData=VTune, FT=TCAS2025.10 | 62.4 | 44.4 | — | 13.4 | — | 34.8 | — | 37.3 | — | |
| TimeChatData=TimeIT, FT=SFT2025.10 | 55.8 | 50.9 | — | 10.5 | — | 16.7 | — | 25.7 | — | |
| videollamaData=VTune, FT=SFT2025.10 | 54.4 | 38.2 | — | 10.9 | — | 30.7 | — | 30 | — | |
| Gemini 1.5 Flash#Frames=1 fps2024.11 | 34.6 | 28.2 | 81.4 | 22.7 | 65.6 | 20.5 | 59.3 | 24.5 | 70.8 | |
| gemini-1.5tuning=none2025.10 | 34.6 | 29.7 | — | 24.8 | — | 22.8 | — | 24.5 | — | |
| Qwen2.5vlData=VTune, FT=TCAS2025.10 | 34 | 23 | — | 8.1 | — | 19.6 | — | 18.5 | — | |
| TimeChatLLM=Llama2-7B, #Frames=962024.11 | 30.5 | 25 | 82.1 | 5.6 | 18.5 | 14 | 45.9 | 15.6 | 51.2 | |
| TimeChatData=✗, FT=✗2025.10 | 30.5 | 25 | — | 5.6 | — | 14 | — | 15.6 | — | |
| GPT-4o#Frames=102024.11 | 28.5 | 21.2 | 74.3 | 9.3 | 32.8 | 17.8 | 62.4 | 20.3 | 71.3 | |
| GPT-4otuning=none2025.10 | 28.5 | 21.2 | — | 9.3 | — | 17.8 | — | 20.3 | — | |
| Qwen2.5vlData=VTune, FT=SFT2025.10 | 28.3 | 17.5 | — | 6 | — | 15.1 | — | 14.8 | — | |
| VTimeLLMLLM=Vicuna-7B, #Frames=1002024.11 | 27.3 | 22.7 | 83.2 | 7.3 | 26.9 | 11.9 | 43.7 | 13.6 | 49.8 | |
| VTG-LLMLLM=Llama2-7B, #Frames=962024.11 | 26 | 16.1 | 62.1 | 8.3 | 32 | 6 | 23.1 | 10 | 38.4 | |
| Video-LLaMA2LLM=Mistral-7B, #Frames=82024.11 | 20 | 16.8 | 83.8 | 3.8 | 19 | 10.3 | 51.5 | 10.6 | 52.9 | |
| Video-ChatGPTLLM=Vicuna-7B, #Frames=1002024.11 | 14.4 | 12.8 | 89.2 | 1.3 | 8.8 | 6.5 | 44.8 | 7.2 | 50 | |
| Video-LLaMALLM=Vicuna-7B, #Frames=82024.11 | 14.2 | 10.6 | 74.9 | 5.3 | 37.6 | 7.5 | 53.3 | 7.3 | 51.7 | |
| Video-LLaVALLM=Vicuna-7B, #Frames=82024.11 | 9.4 | 7.6 | 80.8 | 2.8 | 30.3 | 5 | 52.8 | 4.7 | 50 | |
| VideoChat2LLM=Vicuna-7B, #Frames=162024.11 | 7.2 | 5.4 | 74.5 | 1 | 13.7 | 3.8 | 52.1 | 3.6 | 50 |