Rule-discovery on Blicket baseline (test)
86Task SuccessQwen3-32B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-32BPrompt=TiO2026.04 | 86 | 12.98 | 1.31 | |
| Qwen3-32B + 32B TiOPrompt=Base, Training=Distilled with Wason Task2026.04 | 77 | 11.02 | 0.88 | |
| Qwen3-32BPrompt=Base2026.04 | 57 | 12.18 | 0.54 | |
| Qwen3-8BPrompt=TiO2026.04 | 40 | 10.73 | 0.57 | |
| Qwen3-8B + 32B TiOPrompt=Base, Training=Distilled with Wason Task2026.04 | 38 | 11.84 | 0.51 | |
| Qwen3-8B + 32BPrompt=Base, Training=Distilled with Wason Task2026.04 | 36 | 10.25 | 0.53 | |
| Qwen3-8B + 8B TiOPrompt=Base, Training=Distilled with Wason Task2026.04 | 34 | 13.96 | 0.48 | |
| Qwen3-8BPrompt=Base2026.04 | 32 | 10.26 | 0.44 |