用多智能体协作推荐，让旅游推荐更公平多样
Collab-REC: An LLM-based Agentic Framework for Balancing Recommendations in Tourism
- 三个大模型智能体分别从个性化、热门度和可持续性角度提建议
- 通过非大模型主持人迭代优化，提升推荐多样性与整体相关性
- 适合关注公平推荐和小众景点挖掘的从业者与研究者
我们提出COLLAB-REC，一种基于大模型的多智能体框架，用于缓解旅游推荐中的热门偏差并提升多样性。系统包含三个基于大模型的智能体（个性化、热门度、可持续性），分别从不同视角生成城市推荐；一名非大模型主持人通过迭代约束优化整合并精炼这些提案，确保各方观点被保留的同时减少冗余输出。在使用不同规模和家族的大模型对欧洲城市查询进行的大量离线实验中，相较于单智能体基线，COLLAB-REC显著提升了推荐的多样性与整体相关性，并有效揭示了常被忽视的冷门目的地。该平衡且上下文感知的方法更好地融合了用户与系统层面的多重考量，展示了大模型驱动推荐系统中多方协作的潜力。代码、数据及其他资源见：https://github.com/ashmibanerjee/collab-rec，提示词详见附录。
原文摘要 · Abstract (English)
We propose COLLAB-REC, a multi-agent framework designed to counteract popularity bias and improve diversity in tourism recommendations. In our setup, three LLM-based agents(Personalization, Popularity, and Sustainability) generate city suggestions from different perspectives. A non-LLM moderator then merges and refines these proposals through iterative constrained refinement, ensuring that each agent's viewpoint is represented while reducing spurious or repeated outputs. Extensive offline experiments on European city queries using LLMs of different sizes and model families show that COLLAB-REC improves both diversity and overall relevance compared to a single-agent baseline, while surfacing lesser-visited destinations that are often overlooked. This balanced, context-aware approach better captures a broader range of user and system-level considerations, highlighting the potential of multi-stakeholder collaboration in LLM-driven recommender systems. Code, data, and other artifacts are available here: https://github.com/ashmibanerjee/collab-rec, while the prompts used are included in the appendix.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。