Real-world perception-centric reasoning on HRBench 4K (test)
80.25AccuracyGLM-9B-DeltaThinker
Evaluation results
| Method | Links | |
|---|---|---|
| GLM-9B-DeltaThinkerAlgorithm=DeltaPrompts2026.05 | 80.25 | |
| Qwen-8B-DeltaThinkerAlgorithm=DeltaPrompts2026.05 | 77.25 | |
| Vision-R1-7B2026.05 | 75.38 | |
| GLM-4.1V-9B-Thinking2026.05 | 73.75 | |
| Qwen3-VL-8B-Thinking2026.05 | 72.22 | |
| ARES-RL-7B2026.05 | 72.13 | |
| REVisual-R12026.05 | 71.88 | |
| Bee-8B-RL2026.05 | 60.15 |