Metraipolis › archive › dc_ba91bd7de677
Multi-Harness Reinforcement Learning for Coding Agents: Training Gains and Harness Dependence — arXiv:2609.04518
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04518
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:18:41.452Z
- SHA-256 of the archived text
84dd0519b570542a7e84eeacc9054a94990e8aee01252debdec1949a67d481aa- Stable URL of this page
https://metraipolis.com/documento/dc_ba91bd7de677
How to cite
Multi-Harness Reinforcement Learning for Coding Agents: Training Gains and Harness Dependence — arXiv:2609.04518 — arXiv (en), https://arxiv.org/abs/2609.04518 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ba91bd7de677 (Metraipolis archive dc_ba91bd7de677, sha256 84dd0519b570542a7e84eeacc9054a94990e8aee01252debdec1949a67d481aa)
BibTeX
@misc{dcba91bd7de677,
title = {Multi-Harness Reinforcement Learning for Coding Agents: Training Gains and Harness Dependence — arXiv:2609.04518},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04518},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ba91bd7de677; sha256 84dd0519b570542a7e84eeacc9054a94990e8aee01252debdec1949a67d481aa},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04518. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Recent systems train coding agents with reinforcement learning (RL) on trajectories collected through multiple execution harnesses. However, gains under the training harnesses leave open whether the improvements transfer to unseen interfaces and how they depend on reward normalization. We present a controlled study that measures RL gains under fixed harnesses and compares the same models across harnesses. We train two Qwen3-8B RL variants from the same supervised fine-tuning (SFT) checkpoint using identical replay data. They differ only in whether rewards for each task are normalized within each harness (Within) or across harnesses (Cross). On SWE-bench Verified, we identify three main findings: (i) Multi-harness RL yields modest gains over the SFT baseline under both normalization rules. Across the four training harnesses, the estimated mean improvements are +0.48 and +0.77 percentage points for Within and Cross, respectively. (ii) Cross increases the fraction of trajectories with nonzero advantage but provides no consistent advantage over Within on an unseen harness across training seeds and evaluation budgets; neither rule significantly improves over SFT on this harness. (iii) Performance varies substantially more across harnesses than with RL training. After averaging over model variants, the highest and lowest source-harness solve rates differ by 7.13 percentage points, far exceeding the estimated training gains. Even after multi-harness RL, the observed performance differences across execution interfaces remain substantially larger than either the gains from training or the differences between normalization rules. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.04518 — 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 2026-10-16T22:18:11.355Z 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).