Reasoning over Uncertain Text by Generative Large Language Models

2024-02-14

AI Summary

Large language models (LLMs) exhibit limited capability in explicit probabilistic reasoning—particularly Bayesian inference. To address this, we introduce BLInD, the first fine-grained benchmark dedicated to evaluating Bayesian reasoning in LLMs, and propose a multi-paradigm prompting framework that reformulates probabilistic inference as Python code generation, probabilistic algorithm invocation, and ProbLog-based logical inference. This unified approach jointly models uncertainty, causal structure, and logical constraints. Our method improves average accuracy across multiple state-of-the-art LLMs on BLInD by +28.6% and demonstrates strong cross-domain generalization on causal question-answering tasks. Key contributions are: (1) the first fine-grained Bayesian inference evaluation benchmark; (2) an interpretable, verifiable multi-paradigm probabilistic prompting framework; and (3) a systematic characterization of fundamental limitations of LLMs in probabilistic modeling, including failures in conditional independence reasoning, causal graph construction, and constraint propagation.

Technology Category

• Reasoning under Uncertainty: Probabilistic Inference
• Machine Learning: Probabilistic Circuits and Graphical Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper considers the challenges Large Language Models (LLMs) face when reasoning over text that includes information involving uncertainty explicitly quantified via probability values. This type of reasoning is relevant to a variety of contexts ranging from everyday conversations to medical decision-making. Despite improvements in the mathematical reasoning capabilities of LLMs, they still exhibit significant difficulties when it comes to probabilistic reasoning. To deal with this problem, we introduce the Bayesian Linguistic Inference Dataset (BLInD), a new dataset specifically designed to test the probabilistic reasoning capabilities of LLMs. We use BLInD to find out the limitations of LLMs for tasks involving probabilistic reasoning. In addition, we present several prompting strategies that map the problem to different formal representations, including Python code, probabilistic algorithms, and probabilistic logical programming. We conclude by providing an evaluation of our methods on BLInD and an adaptation of a causal reasoning question-answering dataset. Our empirical results highlight the effectiveness of our proposed strategies for multiple LLMs.

Problem

• Large Language Models
• Probabilistic Reasoning
• Uncertainty Handling

Innovation

• Bayesian Language Inference Dataset
• Probabilistic Reasoning
• Enhancement Strategies