{
  "id": 14904,
  "url": "https://arxiv.org/abs/2607.16955",
  "title": "CADENCE: Closing the Reasoning Gap via Coverage-Adaptive On-Policy Distillation",
  "summary": "On-policy knowledge distillation transfers reasoning from large teachers to compact students, but existing approaches suffer three compounding failure modes: (i) cold-start collapse, where a fresh student assigns near-zero mass to teacher-preferred tokens; (ii) state-agnostic divergence scheduling, where time-only forward/reverse-KL interpolation ignores the student's coverage state; and (iii) binary reward sparsity, where pass/fail signals discard information from partially correct traces. We p",
  "authors": "Satyam Kumar, Saurabh Jha",
  "category": "research",
  "topics": "regulation,children-education",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-17T20:00:00.000Z",
  "fetched_at": "2026-07-31T05:10:57.675Z",
  "source_slug": "hf-daily",
  "source_name": "HuggingFace Daily Papers",
  "source_homepage": "https://huggingface.co/papers",
  "ethics_ai_record_url": "https://ethics.ai/record/14904",
  "original_url": "https://arxiv.org/abs/2607.16955",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}