arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2607.00601cs.CL
收藏

“别说出来!”:语言模型玩禁忌游戏时的约束、合规与沟通

"Don't Say It!": Constraints, Compliance, and Communication when Language Models Play Taboo

Center for Language and Cognition (CLCG), University of Groningen(格罗宁根大学);University of Milano - Bicocca(米兰比可卡大学)

Sara Candussio, Francesca Padovani, Daniel Scalena, Malvina Nissim

AI总结:

研究语言模型在禁忌游戏中如何在严格词汇约束下生成有效描述,通过提示、生成时约束和内部表示操纵干预,评估合规性与沟通效果,发现模型在约束下表现弱于人类。

摘要:

禁忌游戏要求描述一个目标词,同时避免使用一组禁止词,以便其他玩家猜测。这个看似简单的任务结合了严格的词汇约束和沟通有效的描述需求,使其成为研究LLM在推理时如何应对竞争性需求的理想场景。我们评估了两个开放权重模型,在从提示到生成时约束再到内部表示操纵等逐渐深入的生成过程干预条件下。通过禁止词违规检测、LLM作为评判者衡量生成描述成功唤起目标概念的程度(针对人类和机器猜测者),以及检查模型在约束下采用的策略是否与人类玩家一致,来评估其输出。结果表明,对游戏规则的遵守与沟通有效性在不同条件下存在权衡,且模型作为猜测者明显弱于人类,表明在约束下的词汇基础是当前语言模型的一个开放挑战。

英文摘要:

The game of Taboo requires describing a target word without using a set of forbidden words, so that other players can guess it. This deceptively simple task combines strict lexical constraints with the need for communicatively effective descriptions, making it a compelling playground for examining how LLMs navigate competing demands at inference time. We evaluate two open-weight models under conditions that intervene at progressively deeper levels of the generative process, from prompting to generation-time constraints to internal representations manipulations. We assess their outputs through forbidden word violation detection, LLM-as-a-judge measuring the degree to which generated descriptions successfully evoke the target concept for both human and machine guessers, and examining whether the strategies models adopt under constraint align with those of human players. Our results show that compliance with the rules of the game and communicative effectiveness trade off differently across conditions, and that models remain substantially weaker than humans as guessers, suggesting that lexical grounding under constraint is an open challenge for current language models.

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。

↑