MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910215523270656 |
|---|---|
| author | Liu, Yuxin Ye, Ziang Sun, Yueqing Zhu, Mingye Xiao, Jinwei Han, Zhuowen GU, Qi Cai, Xunliang Zhang, Lei |
| author_facet | Liu, Yuxin Ye, Ziang Sun, Yueqing Zhu, Mingye Xiao, Jinwei Han, Zhuowen GU, Qi Cai, Xunliang Zhang, Lei |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_13037 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning Liu, Yuxin Ye, Ziang Sun, Yueqing Zhu, Mingye Xiao, Jinwei Han, Zhuowen GU, Qi Cai, Xunliang Zhang, Lei Artificial Intelligence 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. |
| title | MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.13037 |