HieraMAS: Optimizing Intra-Node LLM Mixtures and Inter-Node Topology for Multi-Agent Systems

TL;DR

HieraMAS introduces a hierarchical framework for multi-agent systems that optimizes intra-node LLM mixtures and inter-node communication topology, significantly improving task performance and efficiency.

Contribution

It presents a novel hierarchical approach combining intra-node LLM mixtures with inter-node topology optimization, addressing credit assignment challenges with a two-stage algorithm.

Findings

Outperforms existing methods on reasoning and coding benchmarks.

Achieves better cost-performance trade-offs.

Effectively manages credit assignment in complex multi-agent configurations.

Abstract

Multi-agent systems (MAS) built on large language models (LLMs) have shown strong performance across many tasks. Most existing approaches improve only one aspect at a time, such as the communication topology, role assignment, or LLM routing, while treating each agent as a single, indivisible unit. This misses the opportunity to use mixtures of LLMs within an agent to strengthen role-specific abilities. We propose HieraMAS, a hierarchical collaboration framework that combines intra-node LLM mixtures with an inter-node communication topology. HieraMAS introduces supernodes, where each functional role is implemented by multiple heterogeneous LLMs using a propose-synthesis structure. Optimizing HieraMAS creates unique credit-assignment challenges: final task performance depends heavily on the underlying LLMs' capabilities, which can lead reinforcement methods to incorrectly reward suboptimal configurations. To address this, we use a two-stage algorithm: (1) multi-level reward attribution, which provides fine-grained feedback at both the node level and the overall system level; (2) graph classification for topology selection, which treats choosing the communication structure as a holistic decision rather than optimizing edges one by one. Experiments on reasoning and coding benchmarks show that HieraMAS substantially outperforms existing methods while also delivering better cost-performance trade-offs.