Mathematically proves that multi-agent planning workflows are decision-theoretically dominated by a centralized Bayes decision maker, setting fundamental limits on agentic emergent behavior.
March 31, 2026
Original Paper
On the Reliability Limits of LLM-Based Multi-Agent Planning
arXiv · 2603.26993
The Takeaway
The study provides a much-needed reality check on the 'agentic hype,' showing that without new external signals, complex multi-agent graphs are merely lossy versions of a single-prompt model. It formalizes the 'reliability limits' of multi-agent systems, shifting the research focus toward information acquisition rather than just complex orchestration.
From the abstract
This technical note studies the reliability limits of LLM-based multi-agent planning as a delegated decision problem. We model the LLM-based multi-agent architecture as a finite acyclic decision network in which multiple stages process shared model-context information, communicate through language interfaces with limited capacity, and may invoke human review. We show that, without new exogenous signals, any delegated network is decision-theoretically dominated by a centralized Bayes decision mak