{
  "id": 3596,
  "url": "https://arxiv.org/abs/2605.28396v1",
  "title": "ADWIN: Adaptive Windows for Horizon-Aware On-Policy Distillation",
  "summary": "On-policy distillation (OPD) transfers reasoning behavior by training a student on teacher feedback along student-generated trajectories, but standard full-rollout training ties every update to a costly completion and can over-allocate supervision to late positions with low marginal value for the current student. We revisit this assumption through the useful supervision horizon: student-induced rollouts can drift from teacher-preferred continuations, while aligned prefixes may already preserve t",
  "authors": "Kun Liang, Chenming Tang, Clive Bai, Weijie Liu, Saiyong Yang, Yunfang Wu",
  "category": "research",
  "topics": "regulation,children-education",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-27T12:33:44.000Z",
  "fetched_at": "2026-07-14T16:30:23.245Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/3596",
  "original_url": "https://arxiv.org/abs/2605.28396v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}