Metraipolis › archive › dc_eec420c8daa7

SHARPO: Segment-Level Credit Assignment for Agentic Reinforcement Learning — arXiv:2610.00838

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.00838
License (as archived)
CC0-1.0
Archived
2026-10-06T12:11:20.039Z
SHA-256 of the archived text
79958996348b38bfe5e17377690134d9c89732646368be4ea98823152fdc3370
Stable URL of this page
https://metraipolis.com/documento/dc_eec420c8daa7

How to cite

SHARPO: Segment-Level Credit Assignment for Agentic Reinforcement Learning — arXiv:2610.00838 — arXiv (en), https://arxiv.org/abs/2610.00838 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eec420c8daa7 (Metraipolis archive dc_eec420c8daa7, sha256 79958996348b38bfe5e17377690134d9c89732646368be4ea98823152fdc3370)
BibTeX
@misc{dceec420c8daa7,
  title = {SHARPO: Segment-Level Credit Assignment for Agentic Reinforcement Learning — arXiv:2610.00838},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.00838},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eec420c8daa7; sha256 79958996348b38bfe5e17377690134d9c89732646368be4ea98823152fdc3370},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.00838. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Agentic reinforcement learning (RL) trains a large language model (LLM) to act over long, multi-step interactions. However, a single localized error can cause task failure, while trajectory-level rewards provide limited guidance for assigning credit to individual decisions. To address this limitation, we introduce Segment-level Hindsight Advantage Reweighting for Policy Optimization (SHARPO), a credit-assignment mechanism that refines Group Relative Policy Optimization (GRPO) at the level of environment-facing segments. Inspired by the existing on-policy self-distillation (OPSD) method, SHARPO computes teacher-student log-probability gaps within each segment and uses the resulting signal to compute a bounded multiplier on the GRPO advantage. This multiplier is shared by all tokens within the segment, allowing credit to vary across different segments. With Qwen2.5-7B-Instruct, SHARPO outperforms existing baselines on the ALFWorld and WebShop benchmarks, including GRPO, SDAR, RLSD, and StepOPSD.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-02 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.00838 — 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 2026-10-06T10:21:28.116Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).