Preprint
Jul 2026
Agentic Neural Architecture Search
This work proposes a mechanism that bridges these two paradigms: an LLM produces a high-quality seed architecture, then decomposes it into a "slotted architecture", a scaffold with named, interchangeable module slots that automatically defines a bounded, task-specific search space for conventional NAS to explore, without manual engineering.
Seokhoon Jeong, Mijung Kim, Taehwan Kim
· 0 citations