先建图再行动，让智能体更高效地理解环境。
MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning
- 先全局探索构建环境先验，再针对性建图，最后基于地图执行任务。
- 在25个游戏环境中，前沿模型在22个中性能从接近零大幅提升。
- 适合需要长时序推理与环境理解的复杂交互任务研究者。
当前交互式大模型智能体依赖目标引导的逐步规划，环境理解是在执行过程中被动获取的。这种时间顺序颠倒导致‘延迟环境感知’：智能体必须通过试错推断环境约束，陷入低效失败循环。受人类功能感知与认知地图理论启发，我们提出‘先建图再行动’范式（MAP），一个可即插即用的框架，将环境理解前置。MAP包含三个阶段：(1) 全局探索，获取环境通用先验；(2) 任务特定建图，构建结构化认知地图；(3) 知识增强执行，基于地图解决问题。实验显示跨基准与大模型均取得一致提升。在ARC-AGI-3上，MAP使前沿模型在25个游戏环境中的22个超越接近零的基线表现。我们进一步提出MAP-2K数据集，包含建图轨迹，训练结果显示其优于专家执行轨迹，表明环境理解比模仿更根本。
原文摘要 · Abstract (English)
Current interactive LLM agents rely on goal-conditioned stepwise planning, where environmental understanding is acquired reactively during execution rather than established beforehand. This temporal inversion leads to Delayed Environmental Perception: agents must infer environmental constraints through trial-and-error, resulting in an Epistemic Bottleneck that traps them in inefficient failure cycles. Inspired by human affordance perception and cognitive map theory, we propose the Map-then-Act Paradigm (MAP), a plug-and-play framework that shifts environment understanding before execution. MAP consists of three stages: (1) Global Exploration, acquiring environment-general priors; (2) Task-Specific Mapping, constructing a structured cognitive map; and (3) Knowledge-Augmented Execution, solving tasks grounded on the map. Experiments show consistent gains across benchmarks and LLMs. On ARC-AGI-3, MAP enables frontier models to surpass near-zero baseline performance in 22 of 25 game environments. We further introduce MAP-2K, a dataset of map-then-act trajectories, and show that training on it outperforms expert execution traces, suggesting that understanding environments is more fundamental than imitation.
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。