Reproduce “From Agents to Edges: Paired Optimization for Multi-Agent Collaboration” 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
Feng Lin, Xiaotian Yu, Yifan Yang, Aijun Li
PACO jointly optimizes sender and receiver instructions plus a shared convention for each handoff, while keeping model weights and workflow fixed. Its search evaluates proposed packages through end-to-end execution. On HotpotQA, PACO reaches 61.26% Joint F1, versus 59.27% for GEPA and 30.00% for vanilla. With its HotpotQA configuration frozen, PACO scores 54.95% Answer F1 on DROP, versus GEPA’s 49.09%, in a fixed four-agent question-answering chain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.