ABSTRAL: Automatic Design of Multi-Agent Systems Through Iterative Refinement and Topology Optimization

TL;DR

ABSTRAL is a novel framework that iteratively refines multi-agent system designs using contrastive trace analysis, enabling transferability, discoverability of roles, and improved efficiency over traditional methods.

Contribution

It introduces a document-based approach for MAS design, demonstrating transferability of design knowledge and the discovery of new roles through contrastive analysis.

Findings

Multi-agent coordination achieves 26% turn efficiency.

Design knowledge transfers effectively across domains.

Contrastive trace analysis discovers new specialist roles.

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 exhausting the limit, yet still improve over single-agent baselines by discovering parallelizable task decompositions. Second, design knowledge encoded in documents transfers: topology reasoning and role templates learned on one domain provide a head start on new domains, with transferred seeds matching coldstart iteration 3 performance in a single iteration. Third, contrastive trace analysis discovers specialist roles absent from any initial design, a capability no prior system demonstrates. On SOPBench (134 bank tasks, deterministic oracle), ABSTRAL reaches 70% validation / 65.96% test pass rate with a GPT-4o backbone. We release the converged documents as inspectable design rationale.