Video Classification on Drive&Act
58.3AccuracyMPO
Evaluation Results
| Method | Links | |
|---|---|---|
| MPO2025.10 | 58.3 | |
| 5-Shotshots=52025.10 | 54.3 | |
| 3-Shotshots=32025.10 | 54.2 | |
| ProTeGi2025.10 | 53 | |
| SEE2025.10 | 51.7 | |
| PE22025.10 | 50.8 | |
| 1-Shotshots=12025.10 | 50.4 | |
| APE2025.10 | 50.3 | |
| Human2025.10 | 47.3 | |
| EvoPrompt2025.10 | 46.7 | |
| OPRO2025.10 | 46.4 | |
| VIOLABackbone=VideoLLaMA3, Evaluation Protocol=Few-shot (20 samples)2026.01 | 37.6 | |
| CoTprompting=Chain-of-Thought2025.10 | 37.2 | |
| VIOLABackbone=Qwen3-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 37.1 | |
| VideoICLBackbone=Qwen3-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 36.9 | |
| RandomBackbone=Qwen3-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 36.8 | |
| RandomBackbone=VideoLLaMA3, Evaluation Protocol=Few-shot (20 samples)2026.01 | 36 | |
| VoteKBackbone=Qwen3-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 35.5 | |
| VoteKBackbone=VideoLLaMA3, Evaluation Protocol=Few-shot (20 samples)2026.01 | 34.5 | |
| VIOLABackbone=LLaVA-Video, Evaluation Protocol=Few-shot (20 samples)2026.01 | 33.3 | |
| VideoICLBackbone=VideoLLaMA3, Evaluation Protocol=Few-shot (20 samples)2026.01 | 31.8 | |
| +Pseudo AnnotationBackbone=VideoLLaMA3, Evaluation Protocol=Few-shot (20 samples)2026.01 | 28.3 | |
| +Pseudo AnnotationBackbone=Qwen3-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 26.6 | |
| VIOLABackbone=Qwen2-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 26.4 | |
| Zero-shotBackbone=LLaVA-Video, Evaluation Protocol=Zero-shot2026.01 | 24.8 | |
| VoteKBackbone=LLaVA-Video, Evaluation Protocol=Few-shot (20 samples)2026.01 | 24 | |
| RandomBackbone=LLaVA-Video, Evaluation Protocol=Few-shot (20 samples)2026.01 | 23.7 | |
| VideoICLBackbone=LLaVA-Video, Evaluation Protocol=Few-shot (20 samples)2026.01 | 22.2 | |
| Zero-shotBackbone=Qwen3-VL, Evaluation Protocol=Zero-shot2026.01 | 22.1 | |
| +Pseudo AnnotationBackbone=LLaVA-Video, Evaluation Protocol=Few-shot (20 samples)2026.01 | 22.1 | |
| Zero-shotBackbone=VideoLLaMA3, Evaluation Protocol=Zero-shot2026.01 | 21.7 | |
| RandomBackbone=Qwen2-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 20.9 | |
| VoteKBackbone=Qwen2-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 18.1 | |
| +Pseudo AnnotationBackbone=Qwen2-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 15.4 | |
| VideoICLBackbone=Qwen2-VL, Evaluation Protocol=Few-shot (20 samples)2026.01 | 11.2 | |
| Zero-shotBackbone=Qwen2-VL, Evaluation Protocol=Zero-shot2026.01 | 7.2 |