| title | MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning |
|---|---|
| authors | Yuxin Liu, Ziang Ye, Yueqing Sun, Mingye Zhu, Jinwei Xiao, Zhuowen Han, Qi GU, Xunliang Cai, Lei Zhang |
| date | 2026-05-13 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.13037 |
| upvotes | 7 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | 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. |
cite
@misc{liu2026map,
title={MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning},
author={Yuxin Liu and Ziang Ye and Yueqing Sun and Mingye Zhu and Jinwei Xiao and Zhuowen Han and Qi GU and Xunliang Cai and Lei Zhang},
year={2026},
eprint={2605.13037},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.13037}
}