Mind Evolution is an evolutionary search strategy proposed by Google DeepMind to improve the reasoning ability of large language models (LLMs).Evolving Deeper LLM Thinking", which aims to expand the utilization of computing resources during reasoning in an innovative way, allowing the model to handle complex problems more effectively.

Mind Evolution uses language models to generate, reorganize, and optimize candidate solutions, thereby evolving a diverse group of solutions. This method avoids formalizing the underlying reasoning problem and only requires a solution evaluator to guide the model to continuously explore and improve during the reasoning process. It not only avoids the complexity of problem formalization, but also demonstrates good scalability and parallelization capabilities, providing new ideas for the future application of large language models (LLMs) in complex tasks.

Experiments show that Mind Evolution significantly outperforms traditional reasoning strategies such as Best-of-N and Sequential Revision in natural language planning tasks. For example, in the TravelPlanner and Natural Plan benchmarks, Mind Evolution's success rate exceeds 98% without the use of a formal solver.