Algorithmic Research Group / Code & Benchmarks

Neural Architecture Search

A recursive reasoning model that ranks architectures by predicted performance and guides search, achieving 8-10x sample efficiency over random search with zero-shot transfer across datasets.