Metraipolis › archive › dc_970e13c67f6b

Persistent Teacher Anchoring for Tool-Using Agents — arXiv:2609.04773

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.04773
License (as archived)
CC0-1.0
Archived
2026-10-16T22:26:37.726Z
SHA-256 of the archived text
14ce2282b173fd96be81e8fa5633da22abaad255bcc3eb3f0de24de261a07405
Stable URL of this page
https://metraipolis.com/documento/dc_970e13c67f6b

How to cite

Persistent Teacher Anchoring for Tool-Using Agents — arXiv:2609.04773 — arXiv (en), https://arxiv.org/abs/2609.04773 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_970e13c67f6b (Metraipolis archive dc_970e13c67f6b, sha256 14ce2282b173fd96be81e8fa5633da22abaad255bcc3eb3f0de24de261a07405)
BibTeX
@misc{dc970e13c67f6b,
  title = {Persistent Teacher Anchoring for Tool-Using Agents — arXiv:2609.04773},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.04773},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_970e13c67f6b; sha256 14ce2282b173fd96be81e8fa5633da22abaad255bcc3eb3f0de24de261a07405},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04773. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Distillation is common in LLM post-training, where on-policy knowledge distillation (OPKD) uses student-generated trajectories to prepare the student for downstream RL. At each state, the student matches a next-token distribution supplied by the teacher. As the rollout enters states the teacher would not visit, the teacher-student distribution gap can accumulate. In tool use, this gap becomes consequential because student-written calls execute before supervision and their observations shape later prefixes. Proposer-verifier generation addresses this drift by letting the teacher decide which student-proposed text is retained during generation. Existing formulations govern text but leave tool execution outside their scope. We propose Persistent Teacher Anchoring (PTA), a student-induced but teacher-committed rollout construction. PTA retains chunk-level verification and adds turn-level commitment, allowing a call to reach the environment only after the teacher has verified the entire turn. Treating verified chunks as atomic generation units, we introduce persistent lookahead, which fills idle rollout capacity by advancing future samples and carrying unfinished ones across student updates under the fixed verifier. Across Search-R1-style retrieval and DeepEyes-style perception RL, applying PTA before downstream RL improves macro best@4 by 2.5 and 2.8 points over OPKD under the same downstream RL budget, while lookahead improves throughput by 24%.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.04773 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T22:18:11.527Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).