Sports Video Understanding on DeepSport (test)
49.89Fine-Grained Recognition AccuracyDeepSport
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeepSportFrames=9.812025.11 | 49.89 | 41.2 | 19.19 | 22.93 | 37.67 | |
| Qwen3-VL-8B-ThinkingFrames=162025.11 | 46.75 | 31.98 | 23.74 | 21.96 | 34.83 | |
| GPT-5Frames=162025.11 | 46.5 | 32.89 | 27.01 | 22.76 | 35.7 | |
| InternVL3.5-14BFrames=162025.11 | 46.38 | 29.1 | 22.64 | 17.59 | 33.01 | |
| Baseline + Sports Curriculum SFTFrames=8.722025.11 | 46.11 | 32.61 | 17.87 | 17.7 | 32.48 | |
| Qwen3-VL-235B-A22B-ThinkingFrames=162025.11 | 44.96 | 31.03 | 28.72 | 25.59 | 35.36 | |
| Qwen3-VL-8B-InstructFrames=162025.11 | 43.82 | 23.49 | 25.52 | 20.77 | 31.66 | |
| Baseline + SFTFrames=102025.11 | 40.67 | 24.38 | 21.4 | 21.07 | 29.91 | |
| Video-R1Frames=162025.11 | 38.21 | 17.76 | 13.73 | 10.16 | 24.06 | |
| T*Frames=162025.11 | 35.79 | 19.49 | 9.57 | 7.56 | 22.3 | |
| VideoITGFrames=162025.11 | 35.39 | 15.25 | 9.64 | 5.96 | 20.85 | |
| SoccerChatFrames=162025.11 | 31.43 | 29.2 | 8.69 | 10.23 | 23.19 | |
| Qwen2.5-VL-7B-Instruct (Baseline)Frames=162025.11 | 28.2 | 13.71 | 9.09 | 3.13 | 16.98 |