[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85464-en":3,"doc-seo-85464-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85464,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Enabling Agents to Communicate Entirely in Latent Space","Natural language limits LLM-based agent communication because rich internal latent states must be downsampled into discrete tokens, creating ambiguity and information loss that can cause multi-agent coordination failures. Inspired by telepathy, Interlat enables inter-agent latent communication by transmitting an agent’s continuous last hidden states as a thought representation, with an additional learned compression process via latent space reasoning. Experiments show Interlat outperforming fine-tuned chain-of-thought prompting and single-agent baselines across heterogeneous models. Further compression accelerates inference up to 24× while maintaining competitive performance.","Enabling Agents to Communicate Entirely in Latent Space  \nZhuoyun Du 1 ,2 ,3†* Runze Wang2† Huiyu Bai2 Zouying Cao4 Xiaoyong Zhu2 Yu Cheng2 Bo Zheng2 Wei Chen 1 Haochao Ying5B  \n1 State Key Lab of CAD&CG, Zhejiang University  \n2Future Living Lab of Alibaba  \n3Zhejiang Key Laboratory of Medical Imaging Artificial Intelligence  \n4 Shanghai Jiao Tong University 5Zhejiang University {duzy, [haochaoying}@zju.edu.cn](haochaoying}@zju.edu.cn) , [yunze.wrz@alibaba-inc.com](yunze.wrz@alibaba-inc.com)  \narXiv :2511 .09 149v 5 [ cs .LG] 13 Jul 2026  \nAbstract  \nWhile natural language is the de facto communication medium for LLM-based agents, it presents a fundamental constraint. The process of downsampling rich, internal latent states into discrete tokens inherently limits the depth and nuance of information that can be transmitted, thereby hindering collaborative problemsolving. Inspired by telepathy, which bypasses symbolic language in communication, we propose Interlat (Inter-agent Latent Space Communication), a paradigm that leverages the continuous last hidden states of an LLM as a representation of its thought for direct communication (termed “latent communication”) . An additional learned compression process further compresses latent communication via latent space reasoning. Experiments demonstrate that Interlat outperforms both fine-tuned chainof-thought (CoT) prompting and single-agent baselines, even across heterogeneous models, promoting more exploratory behavior and enabling genuine utilization of latent information. Further compression not only substantially accelerates inference by up to 24× but also maintains competitive performance through an efficient information-preserving mechanism. We position this work as a feasibility study of entirely latent space inter-agent communication, and our results highlight its potential, offering valuable insights for future research. Our code is available at [https://github.com/XiaoDu](https://github.com/XiaoDu)flying/Interlat.  \n“The limits of my language mean the limits of my world.”  \n—Ludwig Wittgenstein, Tractatus Logico-Philosophicus, §5.6 .  \n1 Introduction  \nLarge language model (LLM)-based agentic systems have emerged as a promising paradigm for  \n†Equal Contribution.  \n∗Work done during an internship at Alibaba Group.  \nBCorresponding Authors.  \nsolving complex tasks by orchestrating multiple agents through natural language communication (Wang et al., 2025, 2024 ; Zhang et al., 2024b ; Tran et al., 2025) . Despite its human readability, natural language imposes fundamental constraints on inter-agent communication. To communicate, an agent must compress its rich, high-dimensional internal states into a sequence of discrete tokens, typically exposing only a single linear message (i.e., a chain of thought (CoT) (Wei et al., 2022) plan) . This downsampling not only discards alternative reasoning paths, but also incurs substantial redundancy, as much of the generated text serves linguistic coherence rather than task-relevant information (Zhang et al., 2024a) . As a result, languagebased communication can be ambiguous and lossy, which has been identified as a major source of coordination failures in multi-agent systems (Chenet al., 2025 ; Cemri et al., 2025) .  \nTo move beyond language space, we explore the direct transmission of internal representations for more precise and information-preserving communication. In multi-agent settings, we refer to this as latent communication. While direct sharing is challenging for humans, which is often depicted in fictions (Liu, 2008), i.e., telepathy, LLM-based agents naturally perform most of their computation in latent space and produce rich hidden states throughout generation, which can be extracted to support direct, expressive communication. Previous hidden-state-based communication methods either rely on one-shot activation grafting (Ramesh and Li, 2025) or remain coupled to language trajectories (Tang et al., 2025), and typically require ","cbCaicFwB1OHTq7Y","https://ap.wps.com/l/cbCaicFwB1OHTq7Y","pdf",11058831,1,24,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"Why does natural language constrain communication between LLM-based agents?\",\"answer\":\"Agents must compress high-dimensional internal states into discrete tokens, which causes loss of nuance and can introduce ambiguity. This downsampling also discards alternative reasoning paths and adds redundancy from language coherence rather than task-relevant content.\"},{\"question\":\"What is Interlat’s core idea for inter-agent communication?\",\"answer\":\"Interlat transmits continuous last-layer hidden states directly between agents instead of sending decoded discrete tokens. This continuous representation is treated as the latent form of an agent’s thoughts for more precise communication.\"},{\"question\":\"How does Interlat achieve efficient communication without losing task-critical information?\",\"answer\":\"Interlat trains an additional reasoning-based compression process that generates compact latent messages autoregressively in latent space. Experiments show latent messages can be compressed to very few tokens while preserving competitive performance and reducing communication latency up to 24×.\"}]",1784203745,60,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"enabling-agents-to-communicate-entirely-in-latent-space","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/enabling-agents-to-communicate-entirely-in-latent-space/85464/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does natural language constrain communication between LLM-based agents?","Question",{"text":75,"@type":76},"Agents must compress high-dimensional internal states into discrete tokens, which causes loss of nuance and can introduce ambiguity. This downsampling also discards alternative reasoning paths and adds redundancy from language coherence rather than task-relevant content.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is Interlat’s core idea for inter-agent communication?",{"text":80,"@type":76},"Interlat transmits continuous last-layer hidden states directly between agents instead of sending decoded discrete tokens. This continuous representation is treated as the latent form of an agent’s thoughts for more precise communication.",{"name":82,"@type":73,"acceptedAnswer":83},"How does Interlat achieve efficient communication without losing task-critical information?",{"text":84,"@type":76},"Interlat trains an additional reasoning-based compression process that generates compact latent messages autoregressively in latent space. Experiments show latent messages can be compressed to very few tokens while preserving competitive performance and reducing communication latency up to 24×.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":28,"slug":108},5,"Comic","comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]