用动态阶段切换和结构化代码生成，让大模型搜索更高效准确
PhaseNAS: Language-Model Driven Architecture Search with Dynamic Phase Adaptation
- 基于实时得分阈值动态调整搜索阶段，避免盲目探索
- 在CIFAR上搜索时间减少86%，精度不降反升
- 适合需要快速生成高性能视觉模型的开发者
神经架构搜索（NAS）面临探索与效率的权衡，尤其在复杂任务中。现有基于大语言模型（LLM）的方法常因静态搜索策略和模糊的架构表示而受限。我们提出PhaseNAS，一种基于LLM的NAS框架，通过实时得分阈值驱动动态阶段转换，并采用结构化架构模板语言实现一致的代码生成。在NAS-Bench-Macro基准上，PhaseNAS持续发现更高精度且排名更优的架构。在图像分类（CIFAR-10/100）任务中，搜索时间最多减少86%的同时保持或提升精度；在目标检测中，可自动生成mAP更高、资源消耗更低的YOLOv8变体。结果表明，PhaseNAS在多种视觉任务中实现了高效、自适应且可泛化的架构搜索。
原文摘要 · Abstract (English)
Neural Architecture Search (NAS) is challenged by the trade-off between search space exploration and efficiency, especially for complex tasks. While recent LLM-based NAS methods have shown promise, they often suffer from static search strategies and ambiguous architecture representations. We propose PhaseNAS, an LLM-based NAS framework with dynamic phase transitions guided by real-time score thresholds and a structured architecture template language for consistent code generation. On the NAS-Bench-Macro benchmark, PhaseNAS consistently discovers architectures with higher accuracy and better rank. For image classification (CIFAR-10/100), PhaseNAS reduces search time by up to 86% while maintaining or improving accuracy. In object detection, it automatically produces YOLOv8 variants with higher mAP and lower resource cost. These results demonstrate that PhaseNAS enables efficient, adaptive, and generalizable NAS across diverse vision tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。