paper-with-me

홈 › Papers

No Attacker Needed: Unintentional Cross-User Contamination in Shared-State LLM Agents

2026-04-01 · Tiankai Yang, Jiate Li, Yi Nian, Shen Dong, Ruiyao Xu, Ryan Rossi, Kaize Ding, Yue Zhao arxiv

LLM-based agents increasingly operate across repeated sessions, maintaining task states to ensure continuity. In many deployments, a single agent serves multiple users within a team or organization, reusing a shared knowledge layer across user identities. This shared persistence expands the failure surface: information that is locally valid for one user can silently degrade another user's outcome when the agent reapplies it without regard for scope. We refer to this failure mode as unintentional cross-user contamination (UCC). Unlike adversarial memory poisoning, UCC requires no attacker; it arises from benign interactions whose scope-bound artifacts persist and are later misapplied. We formalize UCC through a controlled evaluation protocol, introduce a taxonomy of three contamination types, and evaluate the problem in two shared-state mechanisms. Under raw shared state, benign interactions alone produce contamination rates of 57--71%. A write-time sanitization is effective when shared state is conversational, but leaves substantial residual risk when shared state includes executable artifacts, with contamination often manifesting as silent wrong answers. These results indicate that shared-state agents need artifact-level defenses beyond text-level sanitization to prevent silent cross-user failures.

📄 PDF Abstract BibTeX arXiv:2604.01350

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Malicious and Unintentional Disclosure Risks in Large Language Models for Code Generation

2025-03-27 · Rafiqul Rabin, Sean McGregor, Nick Judd

This paper explores the risk that a large language model (LLM) trained for code generation on data mined from software repositories will generate content that discloses sensitive information included in its training data…

Code GenerationLanguage ModelingLanguage ModellingLarge Language Model+1

LNE-Blocking: An Efficient Framework for Contamination Mitigation Evaluation on Large Language Models

2025-09-18 · Ruijie Hou, Yueyang Jiao, Hanxu Hu, Yingming Li 외 arxiv

The problem of data contamination is now almost inevitable during the development of large language models (LLMs), with the training data commonly integrating those evaluation benchmarks even unintentionally. This proble…

MMLU-CF: A Contamination-free Multi-task Language Understanding Benchmark

2024-12-19 · QiHao Zhao, Yangyu Huang, Tengchao Lv, Lei Cui 외

Multiple-choice question (MCQ) datasets like Massive Multitask Language Understanding (MMLU) are widely used to evaluate the commonsense, understanding, and problem-solving abilities of large language models (LLMs). Howe…

MMLUMultiple-choiceMulti-task Language UnderstandingWorld Knowledge

Estimating Contamination via Perplexity: Quantifying Memorisation in Language Model Evaluation

2023-09-19 · Yucheng Li

Data contamination in model evaluation is getting increasingly prevalent as the massive training corpora of large language models often unintentionally include benchmark samples. Therefore, contamination analysis has bec…

Language Model EvaluationLanguage ModelingLanguage ModellingMultiple-choice+1

CAP: Data Contamination Detection via Consistency Amplification

2024-10-19 · Yi Zhao, Jing Li, Linyi Yang

Large language models (LLMs) are widely used, but concerns about data contamination challenge the reliability of LLM evaluations. Existing contamination detection methods are often task-specific or require extra prerequi…