Deductive Video Reasoning on VideoHolmes
45AccuracyGemini-2.5-Pro
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini-2.5-ProSize=-, Evaluation Frames=642026.03 | 45 | — | |
| GPT-4oSize=-, Evaluation Frames=642026.03 | 42 | 57.8 | |
| Gemini-1.5-ProSize=-, Evaluation Frames=642026.03 | 41.2 | — | |
| Insight-V++Size=7B, Evaluation Frames=642026.03 | 40.2 | 54.2 | |
| Insight-VSize=7B, Evaluation Frames=642026.03 | 38.5 | 52.9 | |
| Video-R1Size=7B, Evaluation Frames=642026.03 | 36.5 | — | |
| Qwen2.5-VLSize=7B, Evaluation Frames=642026.03 | 34.7 | 47.3 | |
| VideoChat-R1Size=7B, Evaluation Frames=642026.03 | 33 | — | |
| InternVL2.5Size=8B, Evaluation Frames=642026.03 | 23.6 | — |