Complex Reasoning on TOMATO
38.1AccuracyQwen3-VL-8B + SynRL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL-8B + SynRLParameters=8B, Training=SynRL2026.03 | 38.1 | |
| Qwen3-VL-4B + SynRLParameters=4B, Training=SynRL2026.03 | 36.7 | |
| Qwen3-VL-8BParameters=8B2026.03 | 33.2 | |
| Qwen3-VL-4BParameters=4B2026.03 | 32.1 | |
| VideoLLaMA3-7BParameters=7B2026.03 | 30.1 | |
| LLaVA-OneVision-7BParameters=7B2026.03 | 25.5 | |
| Video-Jigsaw-7BParameters=7B2026.03 | 25.3 | |
| Video-R1-7BParameters=7B2026.03 | 25.1 | |
| VideoLLaMA-2-7BParameters=7B2026.03 | 10.1 |