即使环境部分可观且有随机性，通用智能体仍会学习世界模型。
General Agents Contain World Models, even under Partial Observability and Stochasticity
- 扩展定理至随机与部分可观测环境，证明智能体必然内建世界模型。
- 无需强假设：即使非最优或弱通用智能体也具备环境建模能力。
- 揭示随机化本质是学习环境的必要手段，适用于对智能体认知的研究者。
判断智能体是否拥有对周围环境的模型，是理解其能力与局限性的关键一步。在[10]中，研究证明在特定框架下，几乎所有最优且通用的智能体必然包含足够环境知识，可通过黑箱查询近似重构环境。该结论依赖于智能体确定性及环境完全可观测的假设。本文将该定理推广至随机性与部分可观测环境，表明即使在更现实条件下，随机智能体也无法避免通过随机化学习环境。同时，通过弱化“通用性”定义，进一步证明更弱的智能体同样包含其操作环境的模型。结果揭示随机化本质上是学习环境的必要机制。
原文摘要 · Abstract (English)
Deciding whether an agent possesses a model of its surrounding world is a fundamental step toward understanding its capabilities and limitations. In [10], it was shown that, within a particular framework, every almost optimal and general agent necessarily contains sufficient knowledge of its environment to allow an approximate reconstruction of it by querying the agent as a black box. This result relied on the assumptions that the agent is deterministic and that the environment is fully observable. In this work, we remove both assumptions by extending the theorem to stochastic agents operating in partially observable environments. Fundamentally, this shows that stochastic agents cannot avoid learning their environment through the usage of randomization. We also strengthen the result by weakening the notion of generality, proving that less powerful agents already contain a model of the world in which they operate.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。