Towards a Probabilistic Framework for Analyzing and Improving LLM-Enabled Software

TL;DR

This paper introduces a probabilistic framework to analyze and enhance the reliability of LLM-enabled systems by modeling output distributions, demonstrated through improving autoformalization tasks.

Contribution

It presents a novel probabilistic approach for analyzing and refining LLM-based components, advancing the development of more reliable and interpretable systems.

Findings

Distribution-aware analysis identifies weaknesses in LLM outputs.

Guided alignment improvements enhance system reliability.

Framework facilitates iterative refinement of LLM components.

Abstract

Ensuring the reliability and verifiability of large language model (LLM)-enabled systems remains a significant challenge in software engineering. We propose a probabilistic framework for systematically analyzing and improving these systems by modeling and refining distributions over clusters of semantically equivalent outputs. This framework facilitates the evaluation and iterative improvement of Transference Models--key software components that utilize LLMs to transform inputs into outputs for downstream tasks. To illustrate its utility, we apply the framework to the autoformalization problem, where natural language documentation is transformed into formal program specifications. Our case illustrates how distribution-aware analysis enables the identification of weaknesses and guides focused alignment improvements, resulting in more reliable and interpretable outputs. This principled approach offers a foundation for addressing critical challenges in the development of robust LLM-enabled systems.