发表机构
School of Computer Science, University of Sheffield（谢菲尔德大学计算机科学学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
StateBridge 是一种无训练的潜在通信方法，通过闭式正交变换对齐大语言模型多智能体的隐藏状态，在 26 个模型-任务对中 22 个取得最优或并列最优性能，优于基线。
AI 中文摘要
基于大语言模型的多智能体系统通常以文本形式（即离散 token）进行通信，然而文本会引入离散瓶颈：将发送方的连续隐藏状态转换为离散 token 会丢失仅 token 身份无法捕获的信息。近期研究提出潜在通信作为替代方案，智能体直接传输隐藏表示而无需转换为文本，但现有潜在方法要么在 transformer 中逐层注入工作记忆，要么需要训练好的投影器，限制了可移植性。我们提出 StateBridge，一种无训练的潜在通信方法，通过闭式正交变换将发送方的最后一层隐藏状态对齐到接收方的输入空间，轻量级的归一化校准和词汇锚定确保与预训练输入分布兼容，对齐后的状态作为连续前缀附加到接收方智能体的输入中。我们在数学推理、代码生成和问答任务上，使用两个系列的四个模型评估 StateBridge，结果显示在 26 个模型-任务对中，StateBridge 在 22 个对中取得最佳或并列最佳分数，始终优于最强基线。
英文摘要
Large language model based multi-agent systems usually communicate in text, i.e., using discrete tokens. However, text introduces a discrete bottleneck. Converting the sender's continuous hidden states into discrete tokens discards information that token identities alone cannot capture. Recent work proposes latent communication as an alternative, where agents transmit hidden representations directly without converting them to text. However, existing latent methods either inject working memory layer by layer across the transformers, or require trained projectors that limit portability. We propose StateBridge, a training-free latent communication approach that aligns the sender's final-layer hidden states to the receiver's input space via a closed-form orthogonal transformation. Lightweight norm calibration and vocabulary anchoring ensure compatibility with the pretrained input distribution. The aligned states are prepended to the input of the receiver agent as a continuous prefix. We evaluate StateBridge on math reasoning, code generation, and question answering with four models from two families. StateBridge achieves the best or tied-best score on 22 out of 26 model-task pairs, consistently outperforming the strongest baseline.
Comments18 pages, 3 figures, 4 tables, accepted by COLM2026
