Sports Video Question Answering on SportsTime 200 stratified samples (test)
30.5Human Average ScoreCoTR (full)
Evaluation Results
| Method | Links | |
|---|---|---|
| CoTR (full)training=full, Backbone=Qwen3-VL-4B-Instruct2026.04 | 30.5 | |
| CoTR (tr-GRPO)training=tr-GRPO, Backbone=Qwen3-VL-4B-Instruct2026.04 | 29 | |
| InternVL2.5-4B2026.04 | 26 | |
| Qwen3-VL-4B-Instruct2026.04 | 24.5 | |
| Qwen2.5-VL-3B-Instruct2026.04 | 17.5 | |
| MiniCPM-v4-4B2026.04 | 16.5 | |
| LLaVA-OneVision1.5-4B2026.04 | 12.5 | |
| Ovis2-4B2026.04 | 8 |