Thinking-with-images on HRBench 8K
0.935AccuracyS1-VL-32B-RL
Evaluation Results
| Method | Links | |
|---|---|---|
| S1-VL-32B-RL2026.06 | 0.935 | |
| S1-VL-32B-SFT2026.06 | 0.851 | |
| S1-Omni-Image2026.06 | 0.85 | |
| Qwen3-VL-235B-A22B-Thinking2026.06 | 0.804 | |
| Qwen3-VL-32B-Thinking2026.06 | 0.77 | |
| GPT-52026.06 | 0.7375 | |
| Thyme-VL2026.06 | 0.72 |