Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BAMAS: Structuring Budget-Aware Multi-Agent Systems

Liming Yang, Junyu Luo, Xuanzhe Liu, Yiling Lou, Zhenpeng Chen

TL;DR

BAMAS tackles budget-aware structuring of LLM-based multi-agent systems by jointly optimizing LLM provisioning under a cost budget via Integer Linear Programming and learning an optimal agent collaboration topology through offline reinforcement learning. The approach ensures strong cost-performance trade-offs by selecting a high-performing yet budget-feasible pool of LLMs and an effective collaboration topology that adapts to task type and resource availability. Key findings show BAMAS achieves comparable accuracy to state-of-the-art methods while reducing cost up to 86%, with robust budget adherence and topology adaptation (e.g., favoring simple topologies at low budgets and more complex ones when funds permit). The work contributes a practical framework and release-ready code for budget-aware multi-agent structuring, enabling scalable deployment in cost-constrained LLM systems.

Abstract

Large language model (LLM)-based multi-agent systems have emerged as a powerful paradigm for enabling autonomous agents to solve complex tasks. As these systems scale in complexity, cost becomes an important consideration for practical deployment. However, existing work rarely addresses how to structure multi-agent systems under explicit budget constraints. In this paper, we propose BAMAS, a novel approach for building multi-agent systems with budget awareness. BAMAS first selects an optimal set of LLMs by formulating and solving an Integer Linear Programming problem that balances performance and cost. It then determines how these LLMs should collaborate by leveraging a reinforcement learning-based method to select the interaction topology. Finally, the system is instantiated and executed based on the selected agents and their collaboration topology. We evaluate BAMAS on three representative tasks and compare it with state-of-the-art agent construction methods. Results show that BAMAS achieves comparable performance while reducing cost by up to 86%.

BAMAS: Structuring Budget-Aware Multi-Agent Systems

TL;DR

BAMAS tackles budget-aware structuring of LLM-based multi-agent systems by jointly optimizing LLM provisioning under a cost budget via Integer Linear Programming and learning an optimal agent collaboration topology through offline reinforcement learning. The approach ensures strong cost-performance trade-offs by selecting a high-performing yet budget-feasible pool of LLMs and an effective collaboration topology that adapts to task type and resource availability. Key findings show BAMAS achieves comparable accuracy to state-of-the-art methods while reducing cost up to 86%, with robust budget adherence and topology adaptation (e.g., favoring simple topologies at low budgets and more complex ones when funds permit). The work contributes a practical framework and release-ready code for budget-aware multi-agent structuring, enabling scalable deployment in cost-constrained LLM systems.

Abstract

Large language model (LLM)-based multi-agent systems have emerged as a powerful paradigm for enabling autonomous agents to solve complex tasks. As these systems scale in complexity, cost becomes an important consideration for practical deployment. However, existing work rarely addresses how to structure multi-agent systems under explicit budget constraints. In this paper, we propose BAMAS, a novel approach for building multi-agent systems with budget awareness. BAMAS first selects an optimal set of LLMs by formulating and solving an Integer Linear Programming problem that balances performance and cost. It then determines how these LLMs should collaborate by leveraging a reinforcement learning-based method to select the interaction topology. Finally, the system is instantiated and executed based on the selected agents and their collaboration topology. We evaluate BAMAS on three representative tasks and compare it with state-of-the-art agent construction methods. Results show that BAMAS achieves comparable performance while reducing cost by up to 86%.

Paper Structure

This paper contains 40 sections, 6 equations, 3 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Budget-Constrained LLM Provisioning
  5. Agent Collaboration Topology Selection
  6. Composite Reward Function
  7. Learning Algorithm and Loss Function
  8. Agent Instantiation
  9. Evaluation Setup
  10. Research Questions (RQs)
  11. Datasets
  12. Available LLMs
  13. Evaluation Metrics
  14. Baselines
  15. Configuration of BAMAS
  16. ...and 25 more sections

Figures (3)

  • Figure 1: Overview of BAMAS, which constructs a budget-aware multi-agent system by provisioning a cost-optimal set of LLMs and selecting the best collaboration topology to guide task execution.
  • Figure 2: Average cost and accuracy achieved by BAMAS and baseline approaches across datasets.
  • Figure 3: Collaboration topology distributions across datasets and budgets.