{"ID":2882091,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.11836","arxiv_id":"2508.11836","title":"Finite Automata Extraction: Low-data World Model Learning as Programs from Gameplay Video","abstract":"World models are defined as a compressed spatial and temporal learned representation of an environment. The learned representation is typically a neural network, making transfer of the learned environment dynamics and explainability a challenge. In this paper, we propose an approach, Finite Automata Extraction (FAE), that learns a neuro-symbolic world model from gameplay video represented as programs in a novel domain-specific language (DSL): Retro Coder. Compared to prior world model approaches, FAE learns a more precise model of the environment and more general code than prior DSL-based approaches.","short_abstract":"World models are defined as a compressed spatial and temporal learned representation of an environment. The learned representation is typically a neural network, making transfer of the learned environment dynamics and explainability a challenge. In this paper, we propose an approach, Finite Automata Extraction (FAE), t...","url_abs":"https://arxiv.org/abs/2508.11836","url_pdf":"https://arxiv.org/pdf/2508.11836v2","authors":"[\"Dave Goel\",\"Matthew Guzdial\",\"Anurag Sarkar\"]","published":"2025-08-15T23:05:37Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[]","has_code":false}