mlsanityData to 14 Oct 2026

When Goals Leak Across Relationships: How Cached Goal Representations Shape Cross-Client Decisions

arXiv 2610.17365 ↗First posted 12 Oct 2026cs.AIcited by 0

On the map: LLM agents and tool use → Social simulation and multi-agent debate also fits: Agent memory systems

Abstract (arXiv)

As LLM agents take on more autonomous roles, they may interact with multiple clients and counterparties whose goals and incentives conflict. If an agent maintains shared inference state across these interactions, information associated with one relationship may influence decisions made for another. We test OLMo-3-7B-Instruct in a controlled two-client environment with 20 market scenarios, where each client is assigned an opposing objective. With shared conversational memory, the model deviates from the current client's objective in 22% of eligible decisions. We then investigate how these goals are represented in the KV cache. Linear probes show that client-goal ownership becomes highly recoverable in middle layers and is concentrated in particular attention heads. We use these results to identify where to intervene in the cache and test whether these representations affect the model's decisions. Ablating the competitor's cached goal representations at these locations shifts the model toward the current client. The effect is larger for heads identified by probing than for randomly selected heads, and larger when ablating goal tokens than matched non-goal tokens from the same client. Our results show that client-specific goal representations persist in the KV cache and can causally influence decisions made for other clients, highlighting the importance of relationship-specific state isolation in persistent LLM agents.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use. Builds on 20 of its 23 references that are in the map. First posted on arXiv 12 Oct 2026.

Builds on

20 of 23 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Generative Agents: Interactive Simulacra of Human Behavior7 Apr

Other references · 19

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  2. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  3. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  4. 2023MemGPT: Towards LLMs as Operating Systemscited by 1,389
  5. 2016Understanding intermediate layers using linear classifier probescited by 1,353
  6. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,165
  7. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 876
  8. 2019Designing and Interpreting Probes with Control Taskscited by 640
  9. 2024OLMo: Accelerating the Science of Language Modelscited by 613
  10. 2023Towards Best Practices of Activation Patching in Language Models: Metrics and Methodscited by 359
  11. 2023How do Language Models Bind Entities in Context?cited by 108
  12. 2025Latent Collaboration in Multi-Agent Systemscited by 73
7 more
  1. 2025KVCOMM: Online Cross-context KV-cache Communication for Efficient LLM-based Multi-agent Systemscited by 40
  2. 2025Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Contextcited by 26
  3. 2025Shadow in the Cache: Unveiling and Mitigating Privacy Risks of KV-cache in LLM Inferencecited by 21
  4. 2025Moloch's Bargain: Emergent Misalignment When LLMs Compete for Audiencescited by 8
  5. 2026Out of Sight, Not Out of Mind: Unveiling Latent Attack in Latent-based Multi-Agent Systemscited by 4
  6. 2026When Latent Agents Lie: KV-Cache Integrity in Multi-Agent LLM Collaborationcited by 4
  7. 2026CachePrune: Privacy-Aware and Fine-Grained KV Cache Sharing for Efficient LLM Inferencecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell
  2. 2026The Collaboration Tax: How Much LLM Multi-Agent Systems Pay to Coordinate
  3. 2026Aligned but Not Partner-Specific: Distinguishing How Multimodal LLM Agents Succeed in Reference Games Without Human-Like Conventions
  4. 2026Simulated Customers Never Walk Away: Decision Fidelity of LLM User Simulators Measured Against Real Purchase Outcomes
  5. 2025Technical Report: Evaluating Goal Drift in Language Model Agents
  6. 2026A Behavioural and Representational Evaluation of Goal-Directedness in Language Model Agents
  7. 2026Theory of Scene: Breaking the Symmetry Trap in Multi-Agent LLM Coordination
  8. 2026Learning from Others, Acting for You: Cross-User Memory Sharing for LLM Agents