Neural Program Repair by Jointly Learning to Localize and Repair

Authors: Marko Vasic, Aditya Kanade, Petros Maniatis, David Bieber, Rishabh Singh

ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experimental results show that the joint model significantly outperforms an enumerative solution that uses a pointer based model for repair alone. ... In our experimental evaluation, we evaluate three research questions.
Researcher Affiliation Collaboration Marko Vasic1,2, Aditya Kanade1,3, Petros Maniatis1, David Bieber1, Rishabh Singh1 1Google Brain, USA 2University of Texas at Austin, USA 3IISc Bangalore, India
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain an unambiguous statement or link indicating that the source code for the methodology described in this paper is publicly available.
Open Datasets Yes Primarily, we use ETH-Py1502, a public corpus of Git Hub Python files extensively used in the literature (Raychev et al., 2016; Vechev & Yahav, 2016). It consists of 150K Python source files, already partitioned by its publishers into training and test subsets containing 100K and 50K files, respectively. ... 2https://www.sri.inf.ethz.ch/py150 ... Our second dataset, MSR-Var Misuse, is the public portion of the dataset used by Allamanis et al. (2018). ... 3https://aka.ms/iclr18-prog-graphs-dataset
Dataset Splits Yes It consists of 150K Python source files, already partitioned by its publishers into training and test subsets containing 100K and 50K files, respectively. We split the training set into two sets: training (90K) and validation (10K). ... It consists of 25 C# Git Hub projects, split into four partitions: train, validation, seen test, and unseen test, consisting of 3738, 677, 1807, and 1185 files each.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper mentions using LSTM and processing Python and C# files, but does not provide specific version numbers for any software libraries, frameworks, or programming languages used in the implementation of their models.
Experiment Setup No While the paper describes aspects of data generation and loss functions, it does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) or optimizer settings.