Distilled Self-Critique of LLMs with Synthetic Data: a Bayesian Perspective

arXiv:2312.01957v2 Announce Type: replace-cross
Abstract: This paper proposes an interpretation of RLAIF as Bayesian inference by introducing distilled Self-Critique (dSC), which refines the outputs of a LLM through a Gibbs sampler that is later distilled into a fine-tuned model. Only requiring synthetic data, dSC is exercised in experiments regarding safety, sentiment, and privacy control, showing it can be a viable and cheap alternative to align LLMs. Code released at \url{https://github.com/vicgalle/distilled-self-critique}.

arxiv bayesian critique cs.cl cs.lg data llms perspective synthetic synthetic data type