Continuous-state clock reading on Controlled clock benchmark (Combined)
8.2Tolerance Error (1 unit)Qwen3-VL-4B-Instruct
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-VL-4B-Instruct2026.04 | 8.2 | 17.2 | 99.77 | |
| Gemini-3.1-Pro2026.04 | 13.6 | 25.4 | 72.93 | |
| Molmo2-8B2026.04 | 18 | 29.8 | 81.25 | |
| Qwen3-VL-4B + TriSCAMethod Component=TriSCA2026.04 | 45.6 | 58 | 43.62 |