首份系统梳理LLM推荐中公平性问题的综述
Rethinking Fairness in LLM-Based Recommender Systems: A Survey
- 按偏见机制与公平目标双维度梳理研究框架
- 揭示预训练知识、提示词等六大偏见来源
- 适合关注可信推荐系统的研究人员参考
大语言模型正重塑推荐系统，使其具备更强语义理解、生成能力和交互性。但这一转变也带来新的公平性挑战：偏见可能源于预训练知识、提示词设计、生成解释、解码策略及反馈循环。本文系统回顾了基于大语言模型的推荐系统（LLM4Rec）中的公平性研究，从偏见机制与公平目标两个维度组织现有工作，并结构化梳理评估体系与缓解策略。进一步将公平性与可解释性、隐私保护、鲁棒性及可控性等可信性问题关联。据我们所知，这是首个聚焦于LLM4Rec公平性的综述，旨在为未来全面可靠的公平性评估提供基础框架。
原文摘要 · Abstract (English)
Large Language Models (LLMs) are reshaping recommender systems by enabling more semantic, generative, and interactive recommendation pipelines. However, this shift also introduces new fairness challenges, as biases may arise from pretrained knowledge, prompts, generated explanations, decoding strategies, and feedback loops. This survey provides a systematic review of fairness in LLM-based recommender systems (LLM4Rec), organizing existing studies through a two-dimensional view of bias mechanisms and fairness targets, together with a structured overview of the evaluation landscape and mitigation strategies. We further connect fairness with broader trustworthy concerns, including explainability, privacy, robustness, and controllability. To the best of our knowledge, this is the first survey specifically focused on fairness in LLM4Rec, aiming to provide a structured foundation for future research on comprehensive and reliable fairness evaluation in LLM4Rec.
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。