AI RESEARCH
MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning
arXiv CS.AI
•
ArXi:2605.13037v1 Announce Type: new 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.