Search papers, labs, and topics across Lattice.
Auto-Robotist is introduced, a self-evolving LLM agent that distills robot morphology search traces into a natural-language skill library containing structural archetypes, rules, and supporting designs. This agent retrieves skills to condition LLM edits of elite bodies during search, while also updating the library through Add, Diagnose, and Merge operations based on evaluation results. Experiments across seven EvoGym tasks demonstrate that Auto-Robotist improves cold-start search and enables transferable skills to larger design spaces, outperforming a standard Genetic Algorithm.
LLMs can convert expensive robot design evaluations into reusable, auditable design principles by distilling search traces into explicit natural-language skill libraries.
Large language models (LLMs) are increasingly used as proposal generators for evolutionary robot design, yet most loops remain memoryless: simulator results shape the next population but are not preserved as reusable design knowledge. We present Auto-Robotist, a self-evolving LLM agent that distills morphology-search traces into an explicit natural-language skill library. Each skill stores a structural archetype, evidence-grounded positive and negative rules, and the evaluated designs that support them, making design memory inspectable rather than implicit in a population. During search, the agent retrieves skills to condition LLM edits of elite bodies while retaining a Genetic Algorithm (GA) mutation path for exploration; after evaluation, it updates the library through Add, Diagnose, and Merge. Across seven EvoGym tasks spanning locomotion, traversal, and object interaction, Auto-Robotist improves cold-start 5x5 search and transfers learned skills to 10x10 design spaces, where reference-conditioned transfer outperforms GA on every task. These results suggest that LLM agents can convert expensive physical evaluations into reusable, auditable design principles. Our code will be released upon acceptance.
