Visual Reasoning with Tool Use on HRBench 4K
76.8AccuracyThyme-RL-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Thyme-RL-7BBackbone=7B, Training=RL2025.11 | 76.8 | 1 | |
| CodeV-7B-RLBackbone=7B, Training=RL2025.11 | 76.1 | 33.5 | |
| DeepEyes-7BBackbone=7B2025.11 | 73.1 | 24.7 | |
| Pixel-Reasoner-7BBackbone=7B2025.11 | 70.1 | 20.1 |