AI 中文总结
本研究针对隐式多智能体LLM提出因果审计方法，通过受控消息替换揭示隐式通信的作用机制，发现不同规模Qwen模型在不同数据集上的隐式消息效应存在差异，证明总准确率无法反映隐式消息的实际影响。
AI 中文摘要
基于大语言模型（LLM）的多智能体系统（MAS）中的隐式通信会传输连续的内部表征而非文本，但更强的表征能力并不意味着接收方会使用与任务相关的信息。仅靠最终任务性能也无法揭示观测到的效应是否取决于消息的存在、为评估示例生成的内容，或是由其他智能体提供的信息。我们提出一种因果审计方法，该方法在发送方生成的表征进入接收方的边界处应用受控的消息替换。四种消息设置支持对编码的发送方信息、接收方对消息存在与身份的敏感性、特定示例内容的任务价值，以及其他智能体提供的附加价值这五项指标进行测量。我们将该审计方法应用于Qwen3-4B和Qwen3-8B在GSM8K、ARC-C和MATH-500上的隐式中继任务。在GSM8K上，Qwen3-4B的整体性能效应为-1.00个百分点，可分解为其他示例消息保留的-6.17点效应和特定示例内容带来的+5.17点效应；这两个组成部分的方向在8B模型上发生反转。在MATH-500上，Qwen3-4B的15.00点增益包含其他示例消息保留的8.33点和特定示例内容带来的6.67点，而8B模型的增益主要由前一部分构成。自替换对比进一步表明，特定示例内容与其他智能体的价值是不同的。这些结果表明，总准确率无法确定隐式消息如何影响接收方，并推动受控消息比较成为隐式通信的标准评估方法。
英文摘要
Latent communication in large language model (LLM)-based multi-agent systems (MAS) transmits continuous internal representations instead of text, but greater representational capacity does not establish that the receiver uses task-relevant information. End-task performance alone also cannot reveal whether an observed effect depends on message presence, content generated for the evaluated example, or information supplied by a separate agent. We introduce a causal audit that applies controlled message replacements at the boundary where the sender-produced representation enters the receiver. Four message settings support five measurements of encoded sender information, receiver sensitivity to message presence and identity, the task value of example-specific content, and the additional value supplied by a separate agent. We apply the audit to latent relay with Qwen3-4B and Qwen3-8B on GSM8K, ARC-C, and MATH-500. On GSM8K, the Qwen3-4B overall performance effect of -1.00 percentage point decomposes into a -6.17-point effect retained by an other-example message and a +5.17-point effect attributable to example-specific content; both component directions reverse at 8B. On MATH-500, the Qwen3-4B gain of 15.00 points comprises 8.33 points retained by an other-example message and 6.67 points attributable to example-specific content, while the 8B gain is dominated by the former component. Self-substitution comparisons further show that example-specific content and other-agent value are distinct. These results show that aggregate accuracy does not identify how a latent message affects the receiver and motivate controlled message comparisons as a standard evaluation for latent communication.
