Metraipolis › archive › dc_74e309f22332

Learning Perturbation Robust Policies for LLM Agents with Stable Optimization — arXiv:2609.34064

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34064
License (as archived)
CC0-1.0
Archived
2026-10-07T21:02:03.410Z
SHA-256 of the archived text
e7a0792f79500366e32ea04b851bab87032d5e191b8fddc73d9cfeac77b353c4
Stable URL of this page
https://metraipolis.com/documento/dc_74e309f22332

How to cite

Learning Perturbation Robust Policies for LLM Agents with Stable Optimization — arXiv:2609.34064 — arXiv (en), https://arxiv.org/abs/2609.34064 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_74e309f22332 (Metraipolis archive dc_74e309f22332, sha256 e7a0792f79500366e32ea04b851bab87032d5e191b8fddc73d9cfeac77b353c4)
BibTeX
@misc{dc74e309f22332,
  title = {Learning Perturbation Robust Policies for LLM Agents with Stable Optimization — arXiv:2609.34064},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34064},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_74e309f22332; sha256 e7a0792f79500366e32ea04b851bab87032d5e191b8fddc73d9cfeac77b353c4},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34064. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning (RL) has become an effective post-training paradigm for long-horizon large language model (LLM) agents. However, we find that the resulting policies can be sensitive to various policy perturbations, such as hidden-state noise, pruning, and quantization. In this work, we study how to improve perturbation robustness during policy optimization. We first introduce the notion of a perturbation robust policy and analyze conditions under which perturbed policy updates preserve stable monotonic improvement. Based on this analysis, we introduce Stable Perturbation-Robust Policy Optimization (SPrPO), which applies adaptive and sensitivity-aware perturbations during RL training. We evaluate SPrPO on ALFWorld and WebShop and conduct systematic experiments across multiple perturbation types and scales, showing improved perturbation robustness while maintaining stable policy optimization.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34064 — 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 2026-10-07T20:29:02.835Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).