{"ID":2832106,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.06982","arxiv_id":"2512.06982","title":"LLM-Driven Composite Neural Architecture Search for Multi-Source RL State Encoding","abstract":"Designing state encoders for reinforcement learning (RL) with multiple information sources -- such as sensor measurements, time-series signals, image observations, and textual instructions -- remains underexplored and often requires manual design. We formalize this challenge as a problem of composite neural architecture search (NAS), where multiple source-specific modules and a fusion module are jointly optimized. Existing NAS methods overlook useful side information from the intermediate outputs of these modules -- such as their representation quality -- limiting sample efficiency in multi-source RL settings. To address this, we propose an LLM-driven NAS pipeline in which the LLM serves as a neural architecture design agent, leveraging language-model priors and intermediate-output signals to guide sample-efficient search for high-performing composite state encoders. On a mixed-autonomy traffic control task, our approach discovers higher-performing architectures with fewer candidate evaluations than traditional NAS baselines and the LLM-based GENIUS framework.","short_abstract":"Designing state encoders for reinforcement learning (RL) with multiple information sources -- such as sensor measurements, time-series signals, image observations, and textual instructions -- remains underexplored and often requires manual design. We formalize this challenge as a problem of composite neural architectur...","url_abs":"https://arxiv.org/abs/2512.06982","url_pdf":"https://arxiv.org/pdf/2512.06982v2","authors":"[\"Yu Yu\",\"Qian Xie\",\"Nairen Cao\",\"Li Jin\"]","published":"2025-12-07T20:25:07Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"eess.SY\"]","methods":"[\"Reinforcement Learning\",\"Large Language Model\"]","has_code":false}