Reproduce “Self-Evolving Agentic Workflows via Semantic Decomposition of Complex Tasks” 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
Zichuan Fu, Junjia Qi, Xian Wu, Wenlin Zhang, Yimin Deng, Kaifeng Guo, Xiangyu Zhao
SeaFlow presents a framework for bargaining, negotiation, and persuasion. It routes situations to procedures, usually code, rather than requiring model inference at every decision. A separate coding agent revises conditions and procedures from execution records and outcomes between interactions. The system ranked first in GLEE; its Bargaining workflow averaged 0.118 milliseconds per local decision, not end-to-end latency or proven savings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.