Metraipolis › archive › dc_d440451cbec8

Self-Evolving Coding Agents via Harness-Policy Optimization — arXiv:2610.22791

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.22791
License (as archived)
CC0-1.0
Archived
2026-10-19T08:42:58.604Z
SHA-256 of the archived text
25e0e5e14d3b85e1bfefeda35ec90884a9cb23cbd74cc5f19136d09727059dd7
Stable URL of this page
https://metraipolis.com/documento/dc_d440451cbec8

How to cite

Self-Evolving Coding Agents via Harness-Policy Optimization — arXiv:2610.22791 — arXiv (en), https://arxiv.org/abs/2610.22791 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d440451cbec8 (Metraipolis archive dc_d440451cbec8, sha256 25e0e5e14d3b85e1bfefeda35ec90884a9cb23cbd74cc5f19136d09727059dd7)
BibTeX
@misc{dcd440451cbec8,
  title = {Self-Evolving Coding Agents via Harness-Policy Optimization — arXiv:2610.22791},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.22791},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d440451cbec8; sha256 25e0e5e14d3b85e1bfefeda35ec90884a9cb23cbd74cc5f19136d09727059dd7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22791. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large language model (LLM) agents have shown promising potential in coding tasks, and their performance is jointly determined by the policy model and the harness that mediates their interaction with the environment. Indeed, harness and policy are mutually reinforcing and evolving only one component is suboptimal---a harness tuned for one checkpoint may not target the failure modes of an updated policy, while policy learning remains limited to the trajectories elicited by a fixed harness. In light of this, we propose Self-HPO, a recursive self-improving framework for coding agents that alternates harness and policy optimization, using the evolving policy as its own harness proposer. Specifically, in each round, the current policy first analyzes its own failures and proposes new harness candidates (H-step). The policy is then updated through reinforcement learning (RL) under the selected harness (P-step), and hard tasks identified during training form the next harness search set, keeping harness optimization targeted to the policy's evolving failure modes. Across three coding benchmarks, Self-HPO consistently outperforms harness-only and policy-only baselines, improving Qwen3.5-35B-A3B's average resolve rate from 57.8% to 64.8%, with gains generalizing to out-of-distribution settings. Further analysis shows that both optimized components retain gains even when paired with their base counterparts, while their joint gain exceeds the sum of their individual gains. Self-HPO also improves reliability (Pass@1) and expands the capability frontier (Pass@k) over the base agent. Additionally, we show that with a stronger external model (e.g., Opus 4.8) as the harness proposer, the same framework yields a further performance boost.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.22791 — 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 2026-10-19T08:39:10.124Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).