Understanding AI systems' inner workings is critical for ensuring value alignment and safety.
This review explores mechanistic interpretability: reverse engineering the computational …

Modern generative models demonstrate impressive capabilities, likely stemming from an
ability to identify and manipulate abstract concepts underlying their training data. However …

The rapid progress of research aimed at interpreting the inner workings of advanced
language models has highlighted a need for contextualizing the insights gained from years …

Detoxifying multilingual Large Language Models (LLMs) has become crucial due to their
increasing global use. In this work, we explore zero-shot cross-lingual generalization of …

Safety-aligned large language models (LLMs) sometimes falsely refuse pseudo-harmful
prompts, like" how to kill a mosquito," which are actually harmless. Frequent false refusals …

Large language models (LLMs) can often be made to behave in undesirable ways that they
are explicitly fine-tuned not to. For example, the LLM red-teaming literature has produced a …

Safety fine-tuning helps align Large Language Models (LLMs) with human preferences for
their safe deployment. To better understand the underlying factors that make models safe via …

Stakeholders--from model developers to policymakers--seek to minimize the dual-use risks
of large language models (LLMs). An open challenge to this goal is whether technical …

Large language models (LLMs) are vulnerable to adversarial attacks that can elicit harmful
responses. Defending against such attacks remains challenging due to the opacity of …

Conversational large language models are trained to refuse to answer harmful questions.
However, emergent jailbreaking techniques can still elicit unsafe outputs, presenting an …