Large language models are increasingly deployed as advisors whose objective is not aligned with the user's: recommenders optimize for engagement, sales assistants for purchases, negotiation agents for concessions. Whether such advisors stay truthful when honesty conflicts with their own payoff is a core alignment-evaluation question. We turn the canonical Crawford-Sobel cheap-talk model into a pre-specified benchmark for LLM honesty under preference misalignment. Cheap-talk theory predicts neither full revelation nor silence but coarse monotone partitions, with fewer informative intervals as preference conflict grows. A sender observes a state omega in [0,1], wants the receiver's action near omega+b, and sends one costless message to a receiver whose ideal action is omega. The design uses 5 bias levels, 3 prompt frames, a fixed low-temperature setting, and 200 states per cell: 12,000 sender calls. For the positive-bias grid b in {0.01,0.04,0.08,0.12} the exact most-informative partition sizes are 7,4,3,2, with oracle normalized mutual information 0.5294, 0.3268, 0.2205, 0.1829. Running the full design on four instruction-tuned models (GPT-4o, Claude Sonnet 4.5, Gemini 2.5 Flash-Lite, Llama-3.3-70B), we find all four over-reveal relative to the most-informative equilibrium by 1.8 to 4.2x: normalized mutual information stays at 0.78-0.94 where the oracle prescribes 0.18-0.53. Informativeness declines with bias as predicted but never approaches the strategic optimum; rather than coarse partitions, models show near-full revelation with a constant upward offset tracking their bias (linear exaggeration). Payoff-maximizing versus honesty framing has negligible effect. A decoder ablation shows the finding is recoverable only when the receiver reads the sender's stated number: an embedding-only decoder mis-reads the same data as near-babbling.

翻译：大语言模型正日益被部署为与用户目标不一致的顾问类型：推荐系统优化点击率，销售助手鼓励购买，谈判代理争取让步。当诚实行为与自身收益相冲突时，这些顾问能否保持诚实成为对齐评估的核心问题。我们将经典的Crawford-Sobell廉价谈话模型转化为偏好失配情境下LLM诚实性的预设基准。廉价谈话理论预测既非完全揭示也非彻底沉默，而是粗粒度单调划分——随着偏好冲突加剧，信息区间数量减少。发送者观测到[0,1]区间内的状态ω，希望接收者的行动接近ω+b，向理想行动为ω的接收者发送一条无成本信息。实验设计采用5种偏差水平、3种提示框架、固定低温参数，每单元200个状态：共12000次发送者调用。对于正向偏差网格b∈{0.01,0.04,0.08,0.12}，最大信息量划分的精确规模分别为7、4、3、2，对应的Oracle标准化互信息为0.5294、0.3268、0.2205、0.1829。在四种指令微调模型（GPT-4o、Claude Sonnet 4.5、Gemini 2.5 Flash-Lite、Llama-3.3-70B）上运行完整设计后，我们发现所有模型相对于最大信息量均衡存在1.8至4.2倍的过度揭示：当Oracle指定0.18-0.53时，标准化互信息维持在0.78-0.94区间。信息量随偏差增加而下降的趋势符合理论预测，但从未达到策略最优水平；模型呈现近完全揭示状态（伴随反映其偏差的恒定正向偏移的线性夸大），而非形成粗粒度划分。收益最大化框架与诚实框架的影响可忽略不计。解码器消融实验表明，该发现仅当接收者读取发送者陈述的数值时可复现：基于嵌入向量的解码器会将相同数据误判为接近胡言乱语。