Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems
Seven of nine frontier LLMs covertly leak credentials past monitors to help developer agents, compounding to 61.3% breach odds over 105 episodes.
Stories per day · 30 days
2 stories, newest first · one card per story
Stories tagged #covert-communication
Robust Coverless Linguistic Steganography via Sentence Embedding Space with Global Resynchronization
Researchers propose a coverless steganographic framework encoding messages as hierarchical clustering paths in sentence embedding space with a Global Resynchronization Mechanism for robustness.