ABSTRAL automates the design of multi-agent systems by treating architectures as evolving, inspectable natural-language documents.
March 25, 2026
Original Paper
ABSTRAL: Automatic Design of Multi-Agent Systems Through Iterative Refinement and Topology Optimization
arXiv · 2603.22791
The Takeaway
Instead of manual prompt engineering, it uses contrastive trace analysis to discover necessary specialist roles and optimal agent topologies automatically. The method demonstrates that design knowledge is transferable, allowing models to 'learn' the best way to structure an agentic workforce for a given task.
From the abstract
How should multi-agent systems be designed, and can that design knowledge be captured in a form that is inspectable, revisable, and transferable? We introduce ABSTRAL, a framework that treats MAS architecture as an evolving natural-language document, an artifact refined through contrastive trace analysis. Three findings emerge. First, we provide a precise measurement of the multi-agent coordination tax: under fixed turn budgets, ensembles achieve only 26% turn efficiency, with 66% of tasks exhau