Planning on Countdown (Accuracy and Constraint Satisfaction)
100AccuracyDream-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Dream-7BDecoding Strategy=Greedy, Constraint Setting=Ours2026.07 | 100 | — | |
| Dream-7BDecoding Strategy=Sampling, Constraint Setting=Ours2026.07 | 100 | — | |
| LLaDA-8BDecoding Strategy=Greedy, Constraint Setting=Ours2026.07 | 100 | — | |
| LLaDA-8BDecoding Strategy=Sampling, Constraint Setting=Ours2026.07 | 100 | — | |
| LLaDA-8BDecoding Strategy=Greedy, Constraint Setting=Base2026.07 | 53.8 | 53.8 | |
| Dream-7BDecoding Strategy=Greedy, Constraint Setting=Base2026.07 | 53.6 | 53.6 | |
| LLaDA-8BDecoding Strategy=Sampling, Constraint Setting=Base2026.07 | 53.5 | 53.5 | |
| Dream-7BDecoding Strategy=Sampling, Constraint Setting=Base2026.07 | 35.2 | 35.2 |