Silencing the Guardrails: Inference-Time Jailbreaking via Dynamic Contextual Representation Ablation
This work addresses the vulnerability of LLMs to jailbreak attacks, exposing the fragility of current alignment mechanisms and highlighting the need for more robust defenses.
The paper tackled the problem of jailbreaking safety-aligned large language models by proposing Contextual Representation Ablation (CRA), a framework that dynamically silences refusal-inducing activation patterns during inference, and demonstrated that it significantly outperforms baselines across multiple models.
While Large Language Models (LLMs) have achieved remarkable performance, they remain vulnerable to jailbreak attacks that circumvent safety constraints. Existing strategies, ranging from heuristic prompt engineering to computationally intensive optimization, often face significant trade-offs between effectiveness and efficiency. In this work, we propose Contextual Representation Ablation (CRA), a novel inference-time intervention framework designed to dynamically silence model guardrails. Predicated on the geometric insight that refusal behaviors are mediated by specific low-rank subspaces within the model's hidden states, CRA identifies and suppresses these refusal-inducing activation patterns during decoding without requiring expensive parameter updates or training. Empirical evaluation across multiple safety-aligned open-source LLMs demonstrates that CRA significantly outperforms baselines. These results expose the intrinsic fragility of current alignment mechanisms, revealing that safety constraints can be surgically ablated from internal representations, and underscore the urgent need for more robust defenses that secure the model's latent space.