Hugging Face Daily PapersJerry Wang, Haibo Jin, Xiaopeng Yuan1 min readpaperadvanced
ANTMAN: Adaptive Need Tracking for Multi-Agent Navigation in Large Information Spaces
Summary
ANTMAN is an adaptive coordination framework that tracks unresolved information needs via a Need Graph, allowing agents to route work and recover locally as evidence accumulates. Experiments show it scales to 16× larger contexts with only 1.23× coordination overhead, far less than static partition baselines, while keeping answer quality high.
- Need Graph records unresolved queries, evidence, prior attempts, and progress to drive worker selection and routing.
- Separating coordination policy from search interfaces lets the same mechanism work across document QA, long-context scaling, and structured navigation.
- In a 16× context increase, ANTMAN’s coordination grows 1.23× versus >15× for partition‑driven baselines.
- Works even when the heavy lifting is delegated to much smaller worker models, preserving answer quality.
Teams building multi‑agent LLM pipelines or large‑scale retrieval systems should care because ANTMAN cuts coordination cost while maintaining performance.
7/10