Continuous-state clock reading on Controlled clock benchmark Clean
16.8Tol-1Qwen3-VL-4B-Instruct
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-VL-4B-Instruct2026.04 | 16.8 | 40.4 | 53.96 | |
| Gemini-3.1-Pro2026.04 | 22 | 41 | 63.12 | |
| Molmo2-8B2026.04 | 39.4 | 63.6 | 32.31 | |
| Qwen3-VL-4B + TriSCAMethod Component=TriSCA2026.04 | 67.2 | 77.6 | 23.54 |