כתבה
arXiv cs.AI ·
Pass the Baton: Trajectory-Relayed On-Policy Distillation
תקציר מקורי באנגליתarXiv:2607.26057v1 Announce Type: cross Abstract: On-policy distillation (OPD) grounds token-level supervision in the student's own trajectory, yet suffers from prefix failure: once the student commits to a wrong reasoning direction, all subsequent generation builds on this deviation, producing misdirected continuations that elicit unreliable supervision and waste compute. We identify a teacher-student continuation asymmetry on failed prefixes, where the teacher tends to redirect while the student continues along the original direction, and convert it into a label-free handoff trigger in Relay On-Policy Distillation (Relay-OPD). During training, Relay-OPD constructs relay trajectories by letting the teacher briefly take over at detected trigger points to produce a teacher leg, after which
קרא במקור המקורי
arxiv.org
פתח כתבה מקורית