Activation-level consistency training (ACT) yields a robust defense against adaptive jailbreaks in reasoning models by aligning internal activations on clean and wrapped prompts, outperforming output-level variants.
Rapid response: Mitigating LLM jailbreaks with a few examples
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
verdicts
UNVERDICTED 3representative citing papers
Formalizes the jailbreak oracle problem for LLMs and introduces Boa, a two-phase breadth-first then depth-first search system to solve it efficiently.
Disentangled Safety Adapters decouple safety computations from task-optimized LLMs via lightweight adapters, yielding up to 53% better AUC on safety tasks and dynamic inference-time alignment with reduced performance trade-offs.
citing papers explorer
-
Mitigating Adaptive Attacks against Reasoning Models with Activation Consistency Training
Activation-level consistency training (ACT) yields a robust defense against adaptive jailbreaks in reasoning models by aligning internal activations on clean and wrapped prompts, outperforming output-level variants.
-
Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle Problem
Formalizes the jailbreak oracle problem for LLMs and introduces Boa, a two-phase breadth-first then depth-first search system to solve it efficiently.
-
Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment
Disentangled Safety Adapters decouple safety computations from task-optimized LLMs via lightweight adapters, yielding up to 53% better AUC on safety tasks and dynamic inference-time alignment with reduced performance trade-offs.