Reproduce “MAP: Multi-user Personalization with Collaborative LLM-powered Agents” 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
Christine Lee, Jihye Choi, Bilge Mutlu
MAP (Multi-Agent for Multi-User Personalization) is a system of collaborative LLM-powered agents designed to provide adaptive personalization in multi-user environments by accommodating diverse preferences and resolving conflicts. It quantifiably surpasses monolithic LLM approaches in information retrieval and conflict resolution, validating a human-inspired three-stage workflow.
Key insights
- A user-centered workflow inspired by human conflict resolution (Reflection, Analysis, Feedback) provides an effective paradigm for multi-user AI personalization.
- Collaborative multi-agent LLM architectures significantly outperform monolithic LLMs in comprehensive information retrieval and conflict resolution for complex multi-user scenarios.
- Transparency in decision-making and adaptability through user feedback are critical for fostering user trust and enhancing the perceived usability of multi-user AI systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.