Source: Co-STORM-paper

  • Topic: Collaborative STORM (Co-STORM) for complex information seeking.
  • Entities: Yucheng Jiang, Yijia Shao, Dekun Ma, Sina J. Semnani, Monica S. Lam, Co-STORM, WildSeek, STORM, RAG Chatbot, Google Search, You.com, DSPy, Prometheus 2.
  • Concepts: Unknown unknowns, collaborative discourse, multi-agent systems, mind map, information-seeking assistance, mixed-initiative approach, perspective-guided question asking, moderator agent.
  • Key Claims:
    • Co-STORM emulates educational collaborative discourse to help users discover “unknown unknowns” with less mental effort.
    • It uses a multi-agent system (experts + moderator) and a dynamic mind map to track discourse.
    • WildSeek dataset (100 data points, 24 domains) was created for evaluating complex information-seeking.
    • Human evaluation shows 70% preference over search engines and 78% over RAG chatbots.
    • Co-STORM outperforms baselines in depth, novelty, and engagement.