SPLASH 2023
Sun 22 - Fri 27 October 2023 Cascais, Portugal
Tue 24 Oct 2023 11:00 - 11:30 at Room XII - Machine Learning and Synthesis

This paper introduces two methods for automated program repair (APR) utilizing pre-trained language models. The first method demonstrates program repair as a code completion task and is validated on a dataset of Java programs. The second method, Mentat, leverages OCaml’s parser and type system as fault localization techniques to generate prompts for GPT-3, producing candidate patches. Evaluation results show promising repair rates, with 27% and 39.2% effectiveness, respectively. For OCaml, a comparative study employing an automated validation strategy is presented in which the technique outperforms other tools. Language models are effective at APR, enhancing bug fixing and freeing developers to focus on other critical aspects of software engineering.

Tue 24 Oct

Displayed time zone: Lisbon change

11:00 - 12:30
Machine Learning and SynthesisDoctoral Symposium at Room XII
11:00
30m
Talk
Large Language Models for Automated Program Repair
Doctoral Symposium
Francisco Ribeiro University of Minho & HASLab, INESCTEC
11:30
30m
Talk
Scaling up Program Synthesis to Efficient Algorithms
Doctoral Symposium
Ruyi Ji Peking University
12:00
30m
Talk
Transforming Ideas into Code: Visual Sketching for ML Development
Doctoral Symposium
Luis F. Gomes Carnegie Mellon University