mlsanityData to 7 Oct 2026

When Latent Agents Lie: KV-Cache Integrity in Multi-Agent LLM Collaboration

arXiv 2606.28958 ↗First posted 27 Jun 2026cs.MAcited by 2

On the map: LLM agents and tool use → Software engineering coding agents

Abstract (arXiv)

LLM agents can share more than text. In some systems, an agent can send a short visible message while also passing its full KV-cache state to another model. This hidden state can help the final model combine evidence from several agents, but it is also hard to inspect. A visible message may look harmless even if the hidden state has been changed. We study this problem in a multi-agent question-answering setup. Specialists each see part of the evidence, send a short commitment, and pass full KV-cache state to a coordinator. In clean runs, this latent collaboration improves over a matched text-only version. On transformed HiddenBench with Qwen3-4B, it reaches EM/F1 of 0.338/0.486, compared with 0.231/0.369 for text collaboration. Qwen3-8B and HotPotQA runs show the same direction of improvement. The problem appears when one specialist is malicious. Some false visible commitments can steer answers. More seriously, changing the hidden KV state can collapse performance even when the visible commitment still looks plausible. A verifier that checks only text misses this failure mode. Simple magnitude checks catch some obvious corruptions, but adaptive attacks can evade them while still damaging the final answer. The most reliable fix we find is not to guess whether hidden state looks normal, but to protect it in transport. We implement an HMAC-SHA256 manifest that binds the specialist, session, model, visible commitment, tensor metadata, and payload digest. It accepts all 774 honest replayed payloads and rejects all 295 recorded tampered payloads. The main lesson is that full-KV latent memory can be useful, but it should be treated as a security-sensitive object, not as ordinary internal model state.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use. Builds on 34 of its 41 references that are in the map. First posted on arXiv 27 Jun 2026.

Builds on

34 of 41 references.

Closest ancestors · 2

Same direction, newest first

  1. 2026LCGuard: Latent Communication Guard for Safe KV Sharing in Multi-Agent Systems21 May
  2. 2025Why Do Multi-Agent LLM Systems Fail?17 Mar

Other references · 32

  1. 2018Byzantine-Robust Distributed Learning: Towards Optimal Statistical Ratescited by 698
  2. 2024DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Servingcited by 595
  3. 2018The Hidden Vulnerability of Distributed Learning in Byzantiumcited by 324
  4. 2019Robust Aggregation for Federated Learningcited by 314
  5. 2019A Little Is Enough: Circumventing Defenses For Distributed Learningcited by 285
  6. 2024Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systemscited by 126
  7. 2025Latent Collaboration in Multi-Agent Systemscited by 67
  8. 2025Open Challenges in Multi-Agent Security: Towards Secure Systems of Interacting AI Agentscited by 66
  9. 2025Cache-to-Cache: Direct Semantic Communication Between Large Language Modelscited by 54
  10. 2023Let Models Speak Ciphers: Multiagent Debate through Embeddingscited by 49
  11. 2025Language Models are Injective and Hence Invertiblecited by 31
  12. 2025Enabling Agents to Communicate Entirely in Latent Spacecited by 31
20 more
  1. 2025TAMAS: Benchmarking Adversarial Risks in Multi-Agent LLM Systemscited by 24
  2. 2026AgentLeak: A Benchmark for Internal-Channel Privacy Leakage in Multi-Agent LLM Systemscited by 20
  3. 2024C-FedRAG: A Confidential Federated Retrieval-Augmented Generation Systemcited by 19
  4. 2025Systematic Failures in Collective Reasoning under Distributed Information in Multi-Agent LLMscited by 15
  5. 2024Efficient Multi-Agent Collaboration with Tool Use for Online Planning in Complex Table Question Answeringcited by 13
  6. 2025Augmenting Multi-Agent Communication with State Delta Trajectorycited by 13
  7. 2024Information Leakage from Embedding in Large Language Modelscited by 10
  8. 2025An Adversary-Resistant Multi-Agent LLM System via Credibility Scoringcited by 10
  9. 2025MAD-Spear: A Conformity-Driven Prompt Injection Attack on Multi-Agent Debate Systemscited by 10
  10. 2026Latent Space Communication via K-V Cache Alignmentcited by 9
  11. 2026The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systemscited by 8
  12. 2025SentinelNet: Safeguarding Multi-Agent Collaboration Through Credit-Based Dynamic Threat Detectioncited by 7
  13. 2025RAGentA: Multi-Agent Retrieval-Augmented Generation for Attributed Question Answeringcited by 4
  14. 2026HyLaT: Efficient Multi-Agent Communication via Hybrid Latent-Text Protocolcited by 3
  15. 2026Out of Sight, Not Out of Mind: Unveiling Latent Attack in Latent-based Multi-Agent Systemscited by 3
  16. 2026When Embedding-Based Defenses Fail: Rethinking Safety in LLM-Based Multi-Agent Systemscited by 2
  17. 2026Defending LLM-based Multi-Agent Systems Against Cooperative Attacks with Sentence-Level Rectificationcited by 2
  18. 2026Thinking Wrong in Silence: Backdoor Attacks on Continuous Latent Reasoningcited by 1
  19. 2026MASS-RAG: Multi-Agent Synthesis Retrieval-Augmented Generationcited by 1
  20. 2026An Efficient and Privacy-Preserving Architecture for Cross-Institutional Collaborative RAGcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026When Does Latent Communication Pay? A Causal Audit of Relayed KV Caches in Multi-Agent LLMscited by 3
  2. 2026SoK: When Safe Agents Fail Together: The Security of Multi Agent LLM Systemscited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2025KVCOMM: Online Cross-context KV-cache Communication for Efficient LLM-based Multi-agent Systems
  2. 2026Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell
  3. 2026Latent Cache Flow: Model-to-Model Communication Without Text
  4. 2026Beyond tokens: a unified framework for latent communication in LLM-based multi-agent systems
  5. 2026RelayCaching: Accelerating LLM Collaboration via Decoding KV Cache Reuse
  6. 2026StateBridge: Training-free Hidden-state Alignment for Latent Communication in LLM Multi-Agent Systems
  7. 2025Q-KVComm: Efficient Multi-Agent Communication Via Adaptive KV Cache Compression
  8. 2026KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems