Tag
This paper introduces Meta^n, a method for recursive self-improvement in LLM agents by applying a fixed meta-operation to expand reasoning depth, outperforming prior approaches on benchmarks like ARC-AGI-2.