Calibration of Large Language Models on Code Summarization

TL;DR

This paper investigates how to calibrate large language models to reliably assess whether their code summaries are sufficiently similar to human-generated ones, addressing the challenge of evaluating AI-generated summaries without human references.

Contribution

The study introduces methods to predict the likelihood that an LLM-generated code summary resembles a human-written summary, enhancing the reliability of automated code summarization evaluation.

Findings

Approaches can predict the similarity of AI summaries to human summaries.

Calibration methods work across multiple languages and LLMs.

Reliable confidence measures improve trust in AI-generated code summaries.

Abstract

A brief, fluent, and relevant summary can be helpful during program comprehension; however, such a summary does require significant human effort to produce. Often, good summaries are unavailable in software projects, which makes maintenance more difficult. There has been a considerable body of research into automated AI-based methods, using Large Language models (LLMs), to generate summaries of code; there also has been quite a bit of work on ways to measure the performance of such summarization methods, with special attention paid to how closely these AI-generated summaries resemble a summary a human might have produced. Measures such as BERTScore and BLEU have been suggested and evaluated with human-subject studies. However, LLM-generated summaries can be inaccurate, incomplete, etc.: generally, too dissimilar to one that a good developer might write. Given an LLM-generated code summary, how can a user rationally judge if a summary is sufficiently good and reliable? Given just some input source code, and an LLM-generated summary, existing approaches can help judge brevity, fluency and relevance of the summary; however, it's difficult to gauge whether an LLM-generated summary sufficiently resembles what a human might produce, without a "golden" human-produced summary to compare against. We study this resemblance question as calibration problem: given just the code & the summary from an LLM, can we compute a confidence measure, that provides a reliable indication of whether the summary sufficiently resembles what a human would have produced in this situation? We examine this question using several LLMs, for several languages, and in several different settings. Our investigation suggests approaches to provide reliable predictions of the likelihood that an LLM-generated summary would sufficiently resemble a summary a human might write for the same code.