Constraint Satisfaction on Short-generation task
74.3AccuracyICFA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ICFASamples=16 (eff.)2025.12 | 74.3 | 135 | 6.8 | |
| PPO (RLHF)Samples=> 10^5 (train)2025.12 | 68.5 | 140 | 10.2 | |
| Best-of-NN=100, Samples=1002025.12 | 62.1 | 850 | 12 | |
| Beam Searchk=5, Samples=N/A2025.12 | 45.2 | 120 | 18.5 |