环境地图：面向长时间跨度智能体的结构化环境表示

The automation of complex software workflows remains a significant challenge despite the rapid advancements in large language models (LLMs). In long-horizon scenarios, agents often experience cascading errors and environmental stochasticity; a single misstep in a dynamic interface can lead to task failure, resulting in hallucinations or trial-and-error. This research introduces 'Environment Maps', a persistent, agent-agnostic representation that mitigates these failures by consolidating heterogeneous evidence, such as screen recordings and execution traces, into a structured graph. The representation consists of four core components: (1) Contexts (abstracted locations), (2) Actions (parameterized affordances), (3) Workflows (observed trajectories), and (4) Environmental States (dynamic environmental information).