View publication

Existing paradigms for ensuring AI safety, such as guardrail models and alignment training, often compromise either inference efficiency or development flexibility. We introduce Disentangled Safety Adapters (DSA), a novel framework addressing these challenges by decoupling safety-specific computations from a task-optimized base model. DSA utilizes lightweight adapters that leverage the base model's internal representations, enabling diverse and flexible safety functionalities with minimal impact on inference cost. Empirically, DSA-based safety guardrails substantially outperform comparably sized standalone models, notably improving hallucination detection (0.88 vs. 0.61 AUC on Summedits) and also excelling at classifying hate speech (0.98 vs. 0.92 on ToxiGen) and unsafe model inputs and responses (0.93 vs. 0.90 on AEGIS2.0 & BeaverTails). Furthermore, DSA-based safety alignment allows dynamic, inference-time adjustment of alignment strength and a fine-grained trade-off between instruction following performance and model safety. Importantly, combining the DSA safety guardrail with DSA safety alignment facilitates context-dependent alignment strength, boosting safety on StrongReject by 93% while maintaining 98% performance on MTBench -- a total reduction in alignment tax of 8 percentage points compared to standard safety alignment fine-tuning. Overall, DSA presents a promising path towards more modular, efficient, and adaptable AI safety and alignment.

Figure 1: Overview of DSA architecture and how it compares to standard safety techniques.

Related readings and updates.

Mind–body interventions, such as mindfulness-based stress reduction (MBSR), may improve well-being by increasing awareness and regulation of physiological and cognitive states. However, it is unclear how practice may alter long-term, baseline physiological processes, and whether these changes reflect improved well-being. Using respiration rate (RR), which can be sensitive to effects of meditation, and 3 aspects of self-reported well-being...

Read more

The state of the art in machine translation (MT) is governed by neural approaches, which typically provide superior translation accuracy over statistical approaches. However, on the closely related task of word alignment, traditional statistical word alignment models often remain the go-to solution. In this paper, we present an approach to train a Transformer model to produce both accurate translations and alignments. We extract discrete...

Read more