Sakana AI researcher Robert Lange introduced Shinka Evolve, a framework that merges LLMs with evolutionary algorithms for open-ended program search. Unlike systems that optimize solutions to predefined problems, Shinka aims to automatically generate new problems, drawing inspiration from quality-diversity search methods. The system utilizes LLMs as mutation operators and employs a bandit algorithm to adaptively select among frontier models during its search process. Early results show state-of-the-art performance in tasks like circle packing and competitive programming, though its potential as an autonomous scientific researcher is still under development. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON The item describes a new framework and associated research paper for open-ended program evolution using LLMs and evolutionary algorithms.
