Structured Progressive Knowledge Activation for LLM-Driven Neural Architecture Search

2026-04-10

Abstract

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.

Problem

• Neural Architecture Search
• Large Language Models
• Functional Entanglement
• Architecture Knowledge Integration
• Expensive Evaluation

Innovation

• Structured Progressive Knowledge Activation
• Functional Entanglement
• LLM-driven NAS
• Factor-conditioned Editing
• Sample-efficient Architecture Search