{"ID":2896194,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.14172","arxiv_id":"2507.14172","title":"Self-Improving Language Models for Evolutionary Program Synthesis: A Case Study on ARC-AGI","abstract":"Many program synthesis tasks prove too challenging for even state-of-the-art language models to solve in single attempts. Search-based evolutionary methods offer a promising alternative by exploring solution spaces iteratively, but their effectiveness remain limited by the fixed capabilities of the underlying generative model. We propose SOAR, a method that learns program synthesis by integrating language models into a self-improving evolutionary loop. SOAR alternates between (1) an evolutionary search that uses an LLM to sample and refine candidate solutions, and (2) a hindsight learning phase that converts search attempts into valid problem-solution pairs used to fine-tune the LLM's sampling and refinement capabilities\\, -- \\,enabling increasingly effective search in subsequent iterations. On the challenging ARC-AGI benchmark, SOAR achieves significant performance gains across model scales and iterations, leveraging positive transfer between the sampling and refinement finetuning tasks. These improvements carry over to test-time adaptation, enabling SOAR to solve 52\\% of the public test set. Our code is open-sourced at: https://github.com/flowersteam/SOAR","short_abstract":"Many program synthesis tasks prove too challenging for even state-of-the-art language models to solve in single attempts. Search-based evolutionary methods offer a promising alternative by exploring solution spaces iteratively, but their effectiveness remain limited by the fixed capabilities of the underlying generativ...","url_abs":"https://arxiv.org/abs/2507.14172","url_pdf":"https://arxiv.org/pdf/2507.14172v2","authors":"[\"Julien Pourcel\",\"Cédric Colas\",\"Pierre-Yves Oudeyer\"]","published":"2025-07-10T15:42:03Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.AI\",\"cs.NE\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false,"code_links":[{"ID":612250,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2896194,"paper_url":"https://arxiv.org/abs/2507.14172","paper_title":"Self-Improving Language Models for Evolutionary Program Synthesis: A Case Study on ARC-AGI","repo_url":"https://github.com/flowersteam/SOAR","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}