Visual Tool-Use on HRBench 4K
80.1AccuracyARM-Thinker-7B
Evaluation results
| Method | Links | |
|---|---|---|
| ARM-Thinker-7BSize=7B, Backbone=Qwen2.5-VL-7B2025.12 | 80.1 | |
| Mini-o3Source=Lai et al. (2025)2025.12 | 77.5 | |
| Qwen3-VL-8BSize=8B2025.12 | 76.8 | |
| Pixel ReasonerSource=Lai et al. (2025)2025.12 | 74 | |
| DeepEyesSource=Lai et al. (2025)2025.12 | 73.2 | |
| InternVL3-8BSize=8B2025.12 | 70.3 | |
| InternVL3.5-8BSize=8B2025.12 | 69.9 | |
| Qwen2.5-VL-7BSize=7B2025.12 | 69.1 | |
| GPT-4oSource=Lai et al. (2025)2025.12 | 62 |