Metraipolis › archive › dc_d62951fbca1d

A^2Agent: Action-Aware Reinforcement Learning for Repository-Level Code Localization Agents — arXiv:2608.29831

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.29831
License (as archived)
CC0-1.0
Archived
2026-10-17T08:07:09.734Z
SHA-256 of the archived text
b6b08c03a53cce11ae381a9c6811a73ef0b047cf2243476e4408d0cf591efb09
Stable URL of this page
https://metraipolis.com/documento/dc_d62951fbca1d

How to cite

A^2Agent: Action-Aware Reinforcement Learning for Repository-Level Code Localization Agents — arXiv:2608.29831 — arXiv (en), https://arxiv.org/abs/2608.29831 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d62951fbca1d (Metraipolis archive dc_d62951fbca1d, sha256 b6b08c03a53cce11ae381a9c6811a73ef0b047cf2243476e4408d0cf591efb09)
BibTeX
@misc{dcd62951fbca1d,
  title = {A^2Agent: Action-Aware Reinforcement Learning for Repository-Level Code Localization Agents — arXiv:2608.29831},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.29831},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d62951fbca1d; sha256 b6b08c03a53cce11ae381a9c6811a73ef0b047cf2243476e4408d0cf591efb09},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29831. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Localizing issue-relevant code regions is a critical step in automated software engineering. However, due to their reliance on sparse trajectory-level signals, existing methods cannot identify which per-turn actions are effective and often discover correct code regions during exploration but fail to commit them. To address these limitations, we propose an action-aware reinforcement learning method that combines a per-turn reward sequence rewarding both the discovery and commitment of gold code regions with an action-level advantage estimation scheme that isolates each action's credit by grouping turns sharing the same exploration context. Extensive evaluations show that our method improves the average F1 over the state-of-the-art (SOTA) by 1.58% on SWE-Bench Verified and 8.55% on SWE-Bench Pro, with our 4B model outperforming baselines up to 8x larger. Our code is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Software Engineering (cs.SE).
Pagina del lavoro: https://arxiv.org/abs/2608.29831 — 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 2026-10-17T07:47:52.736Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).