Reproduce “Strength Lies in Differences! Improving Strategy Planning for Non-collaborative Dialogues via Diversified User Simulation” 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
Tong Zhang, Chen Huang, Yang Deng, Hongru Liang, Jia Liu, Zujie Wen, Wenqiang Lei, Tat-Seng Chua
A method called TRIP (Tailored Strategic Planning) enhances strategic planning for non-collaborative dialogue agents by integrating a user-aware module and a population-based training paradigm. The approach improves agent adaptability to diverse users, achieving a 0.6888 Success Rate in price negotiation and 0.5533 in charity persuasion, outperforming existing LLM-based baselines.
Key insights
- Explicitly modeling user-specific characteristics (mental states and future actions) through a UASP module enables dialogue agents to devise more tailored and effective strategies.
- Training dialogue agents with a diverse population of user simulators significantly improves their generalization ability and adaptability to various non-collaborative behaviors.
- TRIP's strategy analysis showed lower "Intra-Persona" distances and higher "Inter-Persona" distances, confirming its ability to adapt strategies based on specific user types.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.