An attack framework redirects retrieval-augmented LLM answers using a document set that becomes effective only when all of its parts are retrieved together.
Title resolution pending
5 Pith papers cite this work, alongside 3 external citations. Polarity classification is still indexing.
5
Pith papers citing it
3
external citations · external index
years
2026 5representative citing papers
Manipulating the presentation of market information to LLM pricing agents can shift prices, propagate to rivals, and alter market outcomes without changing the numerical values.
RAG poisoning attacks cause a measurable collapse in document-level attention entropy, which D-SCAN exploits to detect attacks with AUC up to 0.93, including attacks that fail to change the answer.
A single hub text can unreasonably match many images in CLIP-based similarity, exposing vulnerabilities in cross-modal encoders for caption evaluation and retrieval.
citing papers explorer
-
LENS: The Sum Is Worse Than the Parts for Set-Level Poisoning in Retrieval-Augmented Generation
An attack framework redirects retrieval-augmented LLM answers using a document set that becomes effective only when all of its parts are retrieved together.
-
Market Signal Injection: Adversarial Context Manipulation of LLM Pricing Agents
Manipulating the presentation of market information to LLM pricing agents can shift prices, propagate to rivals, and alter market outcomes without changing the numerical values.
-
When Context Bites: Detecting RAG Poisoning via Document-Level Attention Collapse
RAG poisoning attacks cause a measurable collapse in document-level attention entropy, which D-SCAN exploits to detect attacks with AUC up to 0.93, including attacks that fail to change the answer.
-
One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubness
A single hub text can unreasonably match many images in CLIP-based similarity, exposing vulnerabilities in cross-modal encoders for caption evaluation and retrieval.
- Tracing Target Answers in Poisoned Retrieval Corpora via Token Influence Attribution