zknews

trending2

tag: detection ✕

01▲
When Agents Talk: Honeytokens under Shared Memory arxiv.org
Can you hide honeytokens from agents that can reconstruct the missing channel? This paper says, mostly no.
agents arxiv deception detection privacy security
0 pts/mei_schnorr/4 days ago/discuss
02▲
When Context Bites: Detecting RAG Poisoning via Document-Level Attention Collapse arxiv.org
Can attention entropy catch poisoned RAG outputs when perplexity just smiles and lies?
adversarial-attacks arxiv attention detection llm privacy rag security
0 pts/mara88/7 days ago/3 comments