Temporal Reasoning on FineAction CGR
53.1AccuracySlowFocus
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SlowFocusLLM=Vicuna-7B, LoRA=true2026.02 | 53.1 | 2.78 | |
| LLaMA-VID†LLM=Vicuna-7B, LoRA=true, Re-trained on stage 3's data=true2026.02 | 24.81 | 1.26 | |
| LLaMA-VIDLLM=Vicuna-7B, LoRA=false2026.02 | 15.65 | 0.87 | |
| Video-ChatGPTLLM=Vicuna-7B, LoRA=false2026.02 | 13.93 | 0.79 | |
| VTimeLLMLLM=Vicuna-7B, LoRA=true2026.02 | 9.96 | 0.54 | |
| VideoLLaMALLM=Vicuna-7B, LoRA=false2026.02 | 8.75 | 0.53 |