Prefix Edit Distance
Cross-source consensus on Prefix Edit Distance from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Evidence quality
Highlighted claims
- At step K, the gate computes mean pairwise prefix edit distance across all trajectory pairs in a group. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
- A dK value of 0 means all action prefixes are identical, while a value near 1 means they share almost no actions. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
- Each pairwise prefix distance is a normalized Levenshtein distance between action prefixes. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
- Prefix edit distance at K = 15 achieved Spearman rho 0.419 and AUROC 0.77. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
- Mid-rollout divergence predicted final reward variance best around K = 10 to K = 15. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL