Deliberative Alignment: Reasoning Enables Safer Language Models
This addresses safety alignment for language models in critical applications, representing a novel paradigm rather than an incremental improvement.
The paper tackles the challenge of ensuring language models reliably adhere to safety principles in critical domains by introducing Deliberative Alignment, which trains models to explicitly recall and reason over safety specifications before answering. This approach achieved highly precise adherence to OpenAI's safety policies, improving robustness to jailbreaks while decreasing overrefusal rates.
As large-scale language models increasingly impact safety-critical domains, ensuring their reliable adherence to well-defined principles remains a fundamental challenge. We introduce Deliberative Alignment, a new paradigm that directly teaches the model safety specifications and trains it to explicitly recall and accurately reason over the specifications before answering. We used this approach to align OpenAI's o-series models, and achieved highly precise adherence to OpenAI's safety policies, without requiring human-written chain-of-thoughts or answers. Deliberative Alignment pushes the Pareto frontier by simultaneously increasing robustness to jailbreaks while decreasing overrefusal rates, and also improves out-of-distribution generalization. We demonstrate that reasoning over explicitly specified policies enables more scalable, trustworthy, and interpretable alignment.