MetaGen: Self-Evolving Roles and Topologies for Multi-Agent LLM Reasoning

TL;DR

MetaGen tackles the rigidity of fixed role libraries and execution-frozen topologies in multi-agent LLM reasoning by introducing a training-free, inference-time framework that jointly evolves roles and collaboration structure around a minimal backbone. It features an Architect that generates query-conditioned roles, constraint and diversity filters, and a self-evolving DAG orchestration loop with intra-task edits and inter-task memory to reuse verified roles. Empirically, MetaGen delivers higher accuracy with significantly lower token costs across five benchmarks and shows strong robustness to distribution shifts and noise, underscoring practical benefits for scalable, adaptive MAS without backbone weight updates. The work demonstrates that intelligent, auditable, and cost-conscious inference-time orchestration of text-level roles and graphs can outperform both single-agent prompting and fixed-topology baselines in real-world reasoning and code-generation tasks.

Abstract

Large language models are increasingly deployed as multi-agent systems, where specialized roles communicate and collaborate through structured interactions to solve complex tasks that often exceed the capacity of a single agent. However, most existing systems still rely on a fixed role library and an execution-frozen interaction topology, a rigid design choice that frequently leads to task mismatch, prevents timely adaptation when new evidence emerges during reasoning, and further inflates inference cost. We introduce MetaGen, a training-free framework that adapts both the role space and the collaboration topology at inference time, without updating base model weights. MetaGen generates and rewrites query-conditioned role specifications to maintain a controllable dynamic role pool, then instantiates a constrained execution graph around a minimal backbone. During execution, it iteratively updates role prompts and adjusts structural decisions using lightweight feedback signals. Experiments on code generation and multi-step reasoning benchmarks show that MetaGen improves the accuracy and cost tradeoff over strong multi-agent baselines.

Figures (5)

• Figure 1: Overview and positioning of MetaGen. Unlike fixed-role/fixed-topology multi-agent systems and training-based topology designers with execution-frozen graphs, MetaGen enables training-free, query-conditioned role generation and self-evolving topology adjustment entirely at inference time.
• Figure 2: MetaGen framework overview. Given a query, an Architect generates and filters candidate roles, then performs novelty-driven role selection and hybrid graph initialization to form an initial DAG $G_{\text{init}}$. MetaGen supports intra-task evolution by updating role prompts and structure using execution feedback, and inter-task evolution by accumulating cross-instance priors and solidifying verified roles for future reuse.
• Figure 3: Accuracy versus manual prompt size on HumanEval (left) and MMLU (right). Each point corresponds to a different design budget variant, illustrating the trade-off between engineering effort and performance.
• Figure 4: Cold-start recovery after distribution shifts. Accuracy (top) and average tokens (bottom) on the first 20 examples immediately after each shift, comparing Frozen, Random, and MetaGen. MetaGen achieves the strongest cold-start accuracy with lower token cost.
• Figure 5: Robustness to noisy nodes and edges. Left: varying the noise proportion $p$ (fraction of corrupted nodes and optional edges). Right: varying the noise strength level $s$ with fixed $p{=}0.4$.