{"ID":2850972,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.20205","arxiv_id":"2510.20205","title":"Merge and Conquer: Evolutionarily Optimizing AI for 2048","abstract":"Optimizing artificial intelligence (AI) for dynamic environments remains a fundamental challenge in machine learning research. In this paper, we examine evolutionary training methods for optimizing AI to solve the game 2048, a 2D sliding puzzle. 2048, with its mix of strategic gameplay and stochastic elements, presents an ideal playground for studying decision-making, long-term planning, and dynamic adaptation. We implemented two distinct systems: a two-agent metaprompting system where a \"thinker\" large language model (LLM) agent refines gameplay strategies for an \"executor\" LLM agent, and a single-agent system based on refining a value function for a limited Monte Carlo Tree Search. We also experimented with rollback features to avoid performance degradation. Our results demonstrate the potential of evolutionary refinement techniques in improving AI performance in non-deterministic environments. The single-agent system achieved substantial improvements, with an average increase of 473.2 points per cycle, and with clear upward trends (correlation $ρ$=0.607) across training cycles. The LLM's understanding of the game grew as well, shown in its development of increasingly advanced strategies. Conversely, the two-agent system did not garner much improvement, highlighting the inherent limits of meta-prompting.","short_abstract":"Optimizing artificial intelligence (AI) for dynamic environments remains a fundamental challenge in machine learning research. In this paper, we examine evolutionary training methods for optimizing AI to solve the game 2048, a 2D sliding puzzle. 2048, with its mix of strategic gameplay and stochastic elements, presents...","url_abs":"https://arxiv.org/abs/2510.20205","url_pdf":"https://arxiv.org/pdf/2510.20205v1","authors":"[\"Maggie Bai\",\"Ava Kim Cohen\",\"Eleanor Koss\",\"Charlie Lichtenbaum\"]","published":"2025-10-23T04:45:05Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}