Reproduce “DEALS: Decentralized Expertise-Aware Load Serving for Multi-Agent LLM Systems” 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
Jingjuan Huang, Wenbin Wang, Yanchuan Yin, Alvaro Velasquez, Jia Liu
DEALS distributes tasks across language-model agents using local queues and neighbors’ task-specific success rates. Agents weigh workload against past performance when deciding whether to keep or forward a task, without a central controller or LLM call. In tested heterogeneous pools, expertise-aware routing improves accuracy over backlog-only routing; evidence covers complete graphs, three to seven agents, three LLM families, and three benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.