Reproduce “AgentBalance: Backbone-then-Topology Design for Cost-Effective Multi-Agent Systems under Budget Constraints” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shuowei Cai, Yansong Ning, Hao Liu
AgentBalance, developed at The Hong Kong University of Science and Technology, presents a framework for designing cost-effective Multi-Agent Systems (MAS) under explicit token-cost and latency budgets. It achieves superior performance across diverse benchmarks by employing a novel "backbone-then-topology" approach, ensuring practical deployability of LLM-based systems.
Key insights
- For practical web-scale deployment, MAS design must be explicitly "budget-aware," optimizing performance subject to strict token-cost and latency constraints, rather than solely focusing on maximizing performance.
- The choice of LLM backbone is a foundational determinant of the cost-performance frontier for an MAS, often more impactful than modifications to the communication topology.
- Intelligent utilization of heterogeneous LLMs, by matching their varying capabilities and costs to specific agent roles and query complexities, is crucial for achieving cost-effective MAS.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.