State-Dependent Safety Collapse
Autoregressive language models are vulnerable to state-dependent safety collapse via structured multi-turn context manipulation. The vulnerability stems from the model treating dialogue history as a state transition operator rather than a passive record. By initializing a conversational trajectory with a semantic-preserving softened query and a query-aware persona containing specific named entities, an attacker can establish "representational anchors" that trigger abrupt phase transitions in…
Evaluated models: GPT-4o, Claude 3.5 Sonnet, Gemini 2.0 Flash+2 more
- GPT-4o
- Claude 3.5 Sonnet
- Gemini 2.0 Flash
- Llama 3 8B Instruct
- Llama 3 70B Instruct