Reproduce “Beyond Retrieval-Ranking: A Multi-Agent Cognitive Decision Framework for E-Commerce Search” 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
Zhouwei Zhai, Mengxiang Chen, Haoyun Xia, Jin Li, Renquan Zhou, Min Yang
The Multi-Agent Cognitive Decision Framework (MACDF) redefines e-commerce search by replacing the traditional retrieval-ranking paradigm with a collaborative multi-agent system designed to mimic human cognitive decision-making during shopping. Deployed at JD.com, it achieved a 6.5% relative improvement in user conversion rate and a 7.7% reduction in query reformulation count during online A/B testing.
Key insights
- Re-framing e-commerce search as an active "cognitive decision process" rather than passive "information retrieval" aligns system design with users' multi-stage shopping behaviors.
- A collaborative multi-agent system, when tailored with domain-specific knowledge and LLM capabilities, can effectively mimic human problem-solving and provide proactive decision support in e-commerce.
- Explicitly optimizing for the minimization of user decision-making cost, as evidenced by reduced query reformulations, leads to higher user satisfaction and improved business metrics like conversion rates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.