Structured Progressive Knowledge Activation for LLM-Driven Neural Architecture Search

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Liu, Zhen, Liu, Yuhan, Wang, Jinjun, Song, Wei, Liu, Jianyi, Fu, Jingwen
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866917466316210176
author	Liu, Zhen Liu, Yuhan Wang, Jinjun Song, Wei Liu, Jianyi Fu, Jingwen
author_facet	Liu, Zhen Liu, Yuhan Wang, Jinjun Song, Wei Liu, Jianyi Fu, Jingwen
contents	This paper focuses on a key challenge in Neural Architecture Search (NAS): integrating established architectural knowledge while exploring new designs under expensive evaluations. Large language models (LLMs) are a promising assistant for NAS because they can translate rich architectural and coding priors into executable code edits. However, in practice, seemingly local revisions often propagate into non-local behavioral and performance shifts because a single edit can inadvertently couple multiple interacting functional factors, a phenomenon we refer to as functional entanglement. To make LLM knowledge usable under such entanglement, we propose Structured Progressive Knowledge Activation (SPARK), which activates relevant priors by explicitly selecting the functional factor to modify and conditioning the edit on that factor. This factor-conditioned editing reduces entangled side effects and yields more targeted, reliable architecture modifications. On CLRS-DFS, SPARK achieves a 28.1x sample-efficient architecture evolution speedup and yields a 22.9 percent relative improvement in OOD accuracy.
format	Preprint
id	arxiv_https___arxiv_org_abs_2605_04057
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Structured Progressive Knowledge Activation for LLM-Driven Neural Architecture Search Liu, Zhen Liu, Yuhan Wang, Jinjun Song, Wei Liu, Jianyi Fu, Jingwen Machine Learning Artificial Intelligence This paper focuses on a key challenge in Neural Architecture Search (NAS): integrating established architectural knowledge while exploring new designs under expensive evaluations. Large language models (LLMs) are a promising assistant for NAS because they can translate rich architectural and coding priors into executable code edits. However, in practice, seemingly local revisions often propagate into non-local behavioral and performance shifts because a single edit can inadvertently couple multiple interacting functional factors, a phenomenon we refer to as functional entanglement. To make LLM knowledge usable under such entanglement, we propose Structured Progressive Knowledge Activation (SPARK), which activates relevant priors by explicitly selecting the functional factor to modify and conditioning the edit on that factor. This factor-conditioned editing reduces entangled side effects and yields more targeted, reliable architecture modifications. On CLRS-DFS, SPARK achieves a 28.1x sample-efficient architecture evolution speedup and yields a 22.9 percent relative improvement in OOD accuracy.
title	Structured Progressive Knowledge Activation for LLM-Driven Neural Architecture Search
topic	Machine Learning Artificial Intelligence
url	https://arxiv.org/abs/2605.04057