Max Hort

Research shows that errors in natural language can be corrected by translating texts to another language and back using language models. We explore to what extent this latent correction capability extends to Automated Program Repair (APR) by investigating Round-Trip Translation (RTT): translating code from one programming language into another programming or natural language and back, using Large Language Models (LLMs). We hypothesize that RTT restores patterns most commonly seen in the LLM’s training corpora through regression toward the mean, replacing infrequent bugs with more frequent, natural, bug-free code. To test this hypothesis, we employ nine LLMs and four common APR benchmarks in Java, and perform a detailed quantitative and qualitative analysis of RTT-generated patches. We find that RTT through English generates plausible patches for 100 of 164 bugs with GPT-4 on the HumanEval-Java benchmark, and 97 are found to be correct in our manual assessment. Moreover, RTT uniquely generates plausible patches for 46 bugs that were missed by LLMs specifically fine-tuned for APR. While this demonstrates the viability of RTT for APR, we also observe limitations, such as a lower overall bug fix rate than the state-of-the-art and diluting the original coding style. We analyze the impact of these limitations and discuss the potential of using RTT as a complementary component in APR frameworks.

This article provides a comprehensive survey of bias mitigation methods for achieving fairness in Machine Learning (ML) models. We collect a total of 341 publications concerning bias mitigation for ML classifiers. These methods can be distinguished based on their intervention procedure (i.e., pre-processing, in-processing, post-processing) and the technique they apply. We investigate how existing bias mitigation methods are evaluated in the literature. In particular, we consider datasets, metrics, and benchmarking. Based on the gathered insights (e.g., What is the most popular fairness metric? How many datasets are used for evaluating bias mitigation methods?), we hope to support practitioners in making informed choices when developing and evaluating new bias mitigation methods.

The use of modern Natural Language Processing (NLP) techniques has shown to be beneficial for software engineering tasks, such as vulnerability detection and type inference. However, training deep NLP models requires significant computational resources. This paper explores techniques that aim at achieving the best usage of resources and available information in these models. We propose a generic approach, EarlyBIRD, to build composite representations of code from the early layers of a pre-trained transformer model. We empirically investigate the viability of this approach on the CodeBERT model by comparing the performance of 12 strategies for creating composite representations with the standard practice of only using the last encoder layer. Our evaluation on four datasets shows that several early layer combinations yield better performance on defect detection, and some combinations improve multi-class classification. More specifically, we obtain a +2 average improvement of detection accuracy on Devign with only 3 out of 12 layers of CodeBERT and a 3.3x speed-up of fine-tuning. These findings show that early layers can be used to obtain better results using the same resources, as well as to reduce resource usage during fine-tuning and inference.

Fairness is a crucial non-functional requirement of modern software systems that rely on the use of Artificial Intelligence (AI) to make decisions regarding our daily lives in application domains such as justice, healthcare and education. In fact, these algorithms can exhibit unwanted discriminatory behaviours that create unfair outcomes when the software is used, such as giving privilege to one group of users over another (e.g., males vs. females). Mitigating algorithmic bias during the development life cycle of AI-enabled software is crucial given that any bias in these algorithms is inherited by the software systems using them. However, previous work has shown that mitigating bias can impact the performance of such systems. Therefore, we propose herein a novel use of soft computing for improving AI-enabled software fairness. Specifically, we exploit multi-objective search, as opposed to previous work optimising fairness only, to strike an optimal balance between reducing gender bias and improving semantic correctness of word embedding models, which are at the core of many AI-enabled systems. To assess the effectiveness of our proposal, we carry out a thorough empirical study based on the most recent best practice for the evaluation of search-based approaches and AI-enabled software. We explore seven different search-based approaches, and benchmark them against both baseline and state-of-the-art approaches applied to a popular and widely used word embedding model, namely Word2Vec. Our results show that multi-objective search outperforms single-objective search, and generates word embeddings that are strictly better than the original ones in both objectives, bias and semantic correctness, for all investigated cases. Additionally, our approach generates word embeddings of higher semantic correctness than those generated by using state-of-the-art techniques in all cases, while also achieving a higher degree of fairness in 67% of the cases. These findings show the feasibility and effectiveness of multi-objective search as a tool for engineers to incorporate fair and accurate word embedding models in their AI-enabled systems.

The increasingly wide uptake of Machine Learning (ML) has raised the significance of the problem of tackling bias (i.e., unfairness), making it a primary software engineering concern. In this paper, we introduce Fairea, a model behaviour mutation approach to benchmarking ML bias mitigation methods. We also report on a large-scale empirical study to test the effectiveness of 12 widely-studied bias mitigation methods. Our results reveal that, surprisingly, bias mitigation methods have a poor effectiveness in 49% of the cases. In particular, 15% of the mitigation cases have worse fairness-accuracy trade-offs than the baseline established by Fairea; 34% of the cases have a decrease in accuracy and an increase in bias. Fairea has been made publicly available for software engineers and researchers to evaluate their bias mitigation methods.