BIRD: A Trustworthy Bayesian Inference Framework for Large Language Models

arXiv:2404.12494v1 Announce Type: new
Abstract: Large language models primarily rely on inductive reasoning for decision making. This results in unreliable decisions when applied to real-world tasks that often present incomplete contexts and conditions. Thus, accurate probability estimation and appropriate interpretations are required to enhance decision-making reliability. In this paper, we propose a Bayesian inference framework called BIRD for large language models. BIRD provides controllable and interpretable probability estimation for model decisions, based on abductive factors, LLM entailment, as well as …

abstract arxiv bayesian bayesian inference bird cs.cl decision decision making decisions framework inductive inference language language models large language large language models making paper probability reasoning reliability results tasks trustworthy type world