迈向安全且人本对齐的游戏对话推荐的多智能体分解
Toward Safe and Human-Aligned Game Conversational Recommendation via Multi-Agent Decomposition
AI总结：
MATCHA通过多智能体框架提升游戏对话推荐的安全性与人本对齐，实现更精细的个性化和更强的对抗防御能力。
摘要：
对话推荐系统（CRS）随着大语言模型的发展而进步，在电影等领域表现出强大的结果。这些领域通常涉及固定内容和被动消费，用户偏好可以通过类型或主题匹配。相比之下，游戏呈现出不同的挑战：快速演变的目录，由交互驱动的偏好（例如技能水平、机制、硬件），以及在开放对话中产生不安全响应的风险增加。我们提出了MATCHA，一种用于CRS的多智能体框架，该框架分配专门的代理进行意图解析、工具增强检索、多LLM排序与反思、解释和风险控制，从而实现更精细的个性化、长尾覆盖和更强的安全性。在真实用户请求数据集上评估，MATCHA在八个指标上优于六个基线，将Hit@5提高20%，减少流行偏见24%，并实现97.9%的对抗防御。人类和虚拟评委评估确认了解释质量和用户对齐的改善。
英文摘要：
Conversational recommender systems (CRS) have advanced with large language models, showing strong results in domains like movies. These domains typically involve fixed content and passive consumption, where user preferences can be matched by genre or theme. In contrast, games present distinct challenges: fast-evolving catalogs, interaction-driven preferences (e.g., skill level, mechanics, hardware), and increased risk of unsafe responses in open-ended conversation. We propose MATCHA, a multi-agent framework for CRS that assigns specialized agents for intent parsing, tool-augmented retrieval, multi-LLM ranking with reflection, explanation, and risk control which enabling finer personalization, long-tail coverage, and stronger safety. Evaluated on real user request dataset, MATCHA outperforms six baselines across eight metrics, improving Hit@5 by 20%, reducing popularity bias by 24%, and achieving 97.9% adversarial defense. Human and virtual-judge evaluations confirm improved explanation quality and user alignment.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
