CoDocBench: A Dataset for Code-Documentation Alignment in Software Maintenance

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pai, Kunal, Devanbu, Premkumar, Ahmed, Toufique
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913676270764032
author Pai, Kunal
Devanbu, Premkumar
Ahmed, Toufique
author_facet Pai, Kunal
Devanbu, Premkumar
Ahmed, Toufique
contents One of the central tasks in software maintenance is being able to understand and develop code changes. Thus, given a natural language description of the desired new operation of a function, an agent (human or AI) might be asked to generate the set of edits to that function to implement the desired new operation; likewise, given a set of edits to a function, an agent might be asked to generate a changed description, of that function's new workings. Thus, there is an incentive to train a neural model for change-related tasks. Motivated by this, we offer a new, "natural", large dataset of coupled changes to code and documentation mined from actual high-quality GitHub projects, where each sample represents a single commit where the code and the associated docstring were changed together. We present the methodology for gathering the dataset, and some sample, challenging (but realistic) tasks where our dataset provides opportunities for both learning and evaluation. We find that current models (specifically Llama-3.1 405B, Mixtral 8$\times$22B) do find these maintenance-related tasks challenging.
format Preprint
id arxiv_https___arxiv_org_abs_2502_00519
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CoDocBench: A Dataset for Code-Documentation Alignment in Software Maintenance
Pai, Kunal
Devanbu, Premkumar
Ahmed, Toufique
Software Engineering
Machine Learning
One of the central tasks in software maintenance is being able to understand and develop code changes. Thus, given a natural language description of the desired new operation of a function, an agent (human or AI) might be asked to generate the set of edits to that function to implement the desired new operation; likewise, given a set of edits to a function, an agent might be asked to generate a changed description, of that function's new workings. Thus, there is an incentive to train a neural model for change-related tasks. Motivated by this, we offer a new, "natural", large dataset of coupled changes to code and documentation mined from actual high-quality GitHub projects, where each sample represents a single commit where the code and the associated docstring were changed together. We present the methodology for gathering the dataset, and some sample, challenging (but realistic) tasks where our dataset provides opportunities for both learning and evaluation. We find that current models (specifically Llama-3.1 405B, Mixtral 8$\times$22B) do find these maintenance-related tasks challenging.
title CoDocBench: A Dataset for Code-Documentation Alignment in Software Maintenance
topic Software Engineering
Machine Learning
url https://arxiv.org/abs/2502.00519