{
  "id": 4729,
  "url": "https://arxiv.org/abs/2605.07725v1",
  "title": "SOD: Step-wise On-policy Distillation for Small Language Model Agents",
  "summary": "Tool-integrated reasoning (TIR) is difficult to scale to small language models due to instability in long-horizon tool interactions and limited model capacity. While reinforcement learning methods like group relative policy optimization provide only sparse outcome-level rewards. Recently, on-policy distillation (OPD) has gained popularity by supplying dense token-level supervision from a teacher on student-generated trajectories. However, our experiments indicate that applying OPD to TIR leads t",
  "authors": "Qiyong Zhong, Mao Zheng, Mingyang Song, Xin Lin, Jie Sun, Houcheng Jiang et al.",
  "category": "research",
  "topics": "regulation,children-education,agents-autonomy",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-08T13:30:42.000Z",
  "fetched_at": "2026-07-14T16:31:12.746Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4729",
  "original_url": "https://arxiv.org/abs/2605.07725v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}