Multi-agent collaboration helps tasks with sparse steps not tightly linked
Rethinking Multi-Agent Collaboration: When More Is Less
Artificial Intelligence
Summary
As AI agents get smarter, using many agents together doesn't always help and can make things more complicated. The authors found that teamwork among agents works best for tasks that take a long time and have steps that don't depend closely on each other. For tasks where every step closely follows the last, a single agent does better. They created a new system called SAIGE that organizes agent teamwork efficiently by connecting agents based on the meaning of information they share. Their tests showed that adding more agents or deeper collaboration doesn't always improve results.
What this means in practice
- •For autonomous system developers: Design task-focused agent teams for complex, long tasks where steps are loosely connected to improve performance without extra overhead.
- •For workflow automation engineers: Choose single-agent setups for tightly linked, step-by-step workflows to optimize efficiency and avoid unnecessary computational cost.
Authors
Yishuo Yuan, Yibo Wu, Yihan Zhang, Minyuan Sun, Shenliang Li, Xinkai Ma, Yifan Li, Jiaheng Liu
Abstract
The rapid advancement of large language models and single-agent harnesses has reshaped the landscape of autonomous systems, raising a critical question of when multi-agent collaboration offers genuine value. As individual agent capabilities continue to scale, multi-agent collaboration faces diminishing returns while incurring growing context overhead. Through systematic analysis, we delineate the capability boundaries of multi-agent collaboration relative to single-agent alternatives, showing that it confers systematic benefits specifically in long-horizon tasks with sparse dependencies, while single-agent harnesses remain superior in tightly coupled, sequential workflows. Building on these insights, we propose SAIGE, a lightweight multi-agent collaboration mechanism based on Semantic-Aware Incremental Graph Evolution. SAIGE models collaboration as a dynamically evolving graph, where nodes are agent instances spawned on demand and edges encode semantic dependencies established through content-based information retrieval. Experiments on long-horizon, complex task benchmarks show that SAIGE achieves a favorable trade-off between context efficiency and task performance, and that scaling the agent pool or deepening the recursion level does not consistently improve outcomes. Our findings suggest that multi-agent superiority is bounded by task structure rather than universal, and that more agents do not necessarily make a system more intelligent.