Saved in:
Bibliographic Details
Main Authors: Yu, Yu, Xie, Qian, Cao, Nairen, Jin, Li
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.06982
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918244095361024
author Yu, Yu
Xie, Qian
Cao, Nairen
Jin, Li
author_facet Yu, Yu
Xie, Qian
Cao, Nairen
Jin, Li
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06982
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LLM-Driven Composite Neural Architecture Search for Multi-Source RL State Encoding
Yu, Yu
Xie, Qian
Cao, Nairen
Jin, Li
Machine Learning
Systems and Control
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.
title LLM-Driven Composite Neural Architecture Search for Multi-Source RL State Encoding
topic Machine Learning
Systems and Control
url https://arxiv.org/abs/2512.06982