{ "added_at": "2026-10-17", "arxiv_id": "2607.28287", "arxiv_version_reviewed": "v1", "authors": [ "Jens Lehmann", "Andrei Aioanei", "Sahar Vahdati" ], "bibtex_url": "https://awesomerea.org/papers/citations/paper-tycho.bib", "code_url": "https://github.com/NIMI-research/Tycho", "date_basis": "First public arXiv submission date; version v1 was reviewed.", "date_basis_zh": "首次公开提交至 arXiv 的日期；核查版本为 v1。", "date_precision": "day", "description_en": "Tycho constructs editable Python models of game state, transitions, rendering and outcomes from action/observation histories in unfamiliar ARC-AGI-3 games. It checks model predictions against recorded transitions and can use, repair or bypass these models during planning.", "description_zh": "Tycho 根据陌生 ARC-AGI-3 游戏中的动作与观察历史，构建可编辑的 Python 模型，描述游戏状态、状态转移、渲染与结果。它将模型预测与已记录的转移进行核对，并在规划时选择使用、修复或绕过这些模型。", "doi": "10.48550/arXiv.2607.28287", "entry_url": "https://awesomerea.org/papers/paper-tycho/", "entry_url_zh": "https://awesomerea.org/zh/papers/paper-tycho/", "evidence_urls": [ "https://arxiv.org/abs/2607.28287", "https://arxiv.org/html/2607.28287v1", "https://arxiv.org/pdf/2607.28287v1", "https://api.datacite.org/dois/10.48550/arXiv.2607.28287", "https://github.com/NIMI-research/Tycho", "https://raw.githubusercontent.com/NIMI-research/Tycho/f68912a764372ead0a610db2e1c011d41ce5197e/README.md", "https://raw.githubusercontent.com/NIMI-research/Tycho/f68912a764372ead0a610db2e1c011d41ce5197e/LICENSE", "https://raw.githubusercontent.com/NIMI-research/Tycho/f68912a764372ead0a610db2e1c011d41ce5197e/pyproject.toml", "https://raw.githubusercontent.com/NIMI-research/Tycho/f68912a764372ead0a610db2e1c011d41ce5197e/CITATION.cff", "https://raw.githubusercontent.com/NIMI-research/Tycho/f68912a764372ead0a610db2e1c011d41ce5197e/PUBLIC_RELEASE_MANIFEST.json" ], "first_arxiv_date": "2026-07-30", "id": "paper-tycho", "json_url": "https://awesomerea.org/papers/paper-tycho/record.json", "kind": "research", "markdown_url": "https://awesomerea.org/papers/paper-tycho/index.md", "markdown_url_zh": "https://awesomerea.org/zh/papers/paper-tycho/index.md", "metadata_notes": "The arXiv posting date is July 30; the manuscript header and software release use July 29. DataCite confirms the arXiv DOI is registered and findable. The pinned author artifact is Apache-2.0 and includes source, prompts, paper configurations and compact evaluation evidence; ARC environment terms are separate. Python >=3.12, model/ARC access and a Docker or Finch sandbox are documented requirements.", "metadata_notes_zh": "arXiv 发布日期为 7 月 30 日；文稿页首和软件发布采用 7 月 29 日。DataCite 确认 arXiv DOI 已注册且处于可检索状态。固定版本的作者产物采用 Apache-2.0，包含源码、提示词、论文配置和精简评估证据；ARC 环境适用独立条款。文档要求 Python >=3.12、模型与 ARC 访问权限，以及 Docker 或 Finch 沙箱。", "pdf_url": "https://arxiv.org/pdf/2607.28287v1", "project_url": null, "publication_date": "2026-07-30", "publication_status": "preprint", "record_type": "paper", "runtime_tested": false, "scope_limitations_en": "Evaluation uses one stochastic run per configuration on 25 public games and 183 levels; five games informed harness development. Models may abstain on cells or accept bounded rendering variants. Transition replay and game scores do not establish complete engine reconstruction, correct objectives, unseen-state fidelity, unseen-game generalization or run-to-run reliability; the agent can bypass modeling.", "scope_limitations_zh": "评估针对 25 个公开游戏、183 个关卡，每种配置仅进行一次随机运行；其中五个游戏曾用于开发智能体框架。模型可以不预测部分像素格，或接受有限的渲染变体。转移回放和游戏分数不能证明完整引擎重建、目标识别正确、未见状态保真度、未见游戏泛化或多次运行的可靠性；智能体也可以绕过建模。", "targets": [ "game" ], "title": "Tycho: Active Abstraction with Programmatic World Models for ARC-AGI-3", "title_zh": "Tycho：面向 ARC-AGI-3 的程序化世界模型主动抽象", "url": "https://arxiv.org/abs/2607.28287", "venue": "arXiv", "verification": "source-reviewed", "verified_at": "2026-10-17", "why_fits_en": "Studies executable reconstruction of observed game dynamics and checks model fidelity separately from task success, providing a bounded behavioral reconstruction case.", "why_fits_zh": "研究对已观察游戏动态的可执行重建，并将模型保真度与任务成功分开检验，提供了范围明确的行为重建案例。", "year": 2026 }