arXiv subjects

Mark Harman

Publications and source records attributed to Mark Harman.

At least 37 records · Page 2Linked to original sources

arXiv2025

Measuring the Influence of Incorrect Code on Test Generation

It is natural to suppose that a Large Language Model is more likely to generate correct test cases when prompted with correct code under test, compared to incorrect code under test. However, the size of this effect has never been previously measured, despite its obvious importance for both practicing software engineers and researchers. To answer the question, we conducted a comprehensive empirical study on 5 open source and 6 closed source language models, with 3 widely-used benchmark data sets together with 41 repo-level real-world examples from two different real-world data sets. Our results reveal that, when compared to incorrect code under test, LLMs prompted with correct code achieve improvements in test accuracy, code coverage, and bug detection of 57\%, 12\%, and 24\% respectively. We further show that these scientific conclusions carry over from the three benchmark data sets to the real-world code, where tests generated for incorrect code experience a 47\% worse bug detection rate. Finally, we report that improvements of +18\% in accuracy, +4\% coverage, and +34\% in bug detection can be achieved by providing natural language code descriptions. These findings have actionable conclusions. For example, the 47\% reduction in real-world bug detection is a clear concern. Fortunately, it is a concern for which our findings about the added value of descriptions offer an immediately actionable remedy.

cs.SE↗

arXiv2025

Mutation-Guided LLM-based Test Generation at Meta

This paper describes Meta's ACH system for mutation-guided LLM-based test generation. ACH generates relatively few mutants (aka simulated faults), compared to traditional mutation testing. Instead, it focuses on generating currently undetected faults that are specific to an issue of concern. From these currently uncaught faults, ACH generates tests that can catch them, thereby `killing' the mutants and consequently hardening the platform against regressions. We use privacy concerns to illustrate our approach, but ACH can harden code against {\em any} type of regression. In total, ACH was applied to 10,795 Android Kotlin classes in 7 software platforms deployed by Meta, from which it generated 9,095 mutants and 571 privacy-hardening test cases. ACH also deploys an LLM-based equivalent mutant detection agent that achieves a precision of 0.79 and a recall of 0.47 (rising to 0.95 and 0.96 with simple pre-processing). ACH was used by Messenger and WhatsApp test-a-thons where engineers accepted 73% of its tests, judging 36% to privacy relevant. We conclude that ACH hardens code against specific concerns and that, even when its tests do not directly tackle the specific concern, engineers find them useful for their other benefits.

cs.SE↗

arXiv2024

An Empirical Study of the Non-determinism of ChatGPT in Code Generation

There has been a recent explosion of research on Large Language Models (LLMs) for software engineering tasks, in particular code generation. However, results from LLMs can be highly unstable; nondeterministically returning very different codes for the same prompt. Non-determinism is a potential menace to scientific conclusion validity. When non-determinism is high, scientific conclusions simply cannot be relied upon unless researchers change their behaviour to control for it in their empirical analyses. This paper conducts an empirical study to demonstrate that non-determinism is, indeed, high, thereby underlining the need for this behavioural change. We choose to study ChatGPT because it is already highly prevalent in the code generation research literature. We report results from a study of 829 code generation problems from three code generation benchmarks (i.e., CodeContests, APPS, and HumanEval). Our results reveal high degrees of non-determinism: the ratio of coding tasks with zero equal test output across different requests is 75.76%, 51.00%, and 47.56% for CodeContests, APPS, and HumanEval, respectively. In addition, we find that setting the temperature to 0 does not guarantee determinism in code generation, although it indeed brings less non-determinism than the default configuration (temperature=1). These results confirm that there is, currently, a significant threat to scientific conclusion validity. In order to put LLM-based research on firmer scientific foundations, researchers need to take into account non-determinism in drawing their conclusions.

cs.SE↗

arXiv2024

Fairness Improvement with Multiple Protected Attributes: How Far Are We?

Existing research mostly improves the fairness of Machine Learning (ML) software regarding a single protected attribute at a time, but this is unrealistic given that many users have multiple protected attributes. This paper conducts an extensive study of fairness improvement regarding multiple protected attributes, covering 11 state-of-the-art fairness improvement methods. We analyze the effectiveness of these methods with different datasets, metrics, and ML models when considering multiple protected attributes. The results reveal that improving fairness for a single protected attribute can largely decrease fairness regarding unconsidered protected attributes. This decrease is observed in up to 88.3% of scenarios (57.5% on average). More surprisingly, we find little difference in accuracy loss when considering single and multiple protected attributes, indicating that accuracy can be maintained in the multiple-attribute paradigm. However, the effect on F1-score when handling two protected attributes is about twice that of a single attribute. This has important implications for future fairness research: reporting only accuracy as the ML performance metric, which is currently common in the literature, is inadequate.

cs.LG↗

arXiv2024

Enhancing Testing at Meta with Rich-State Simulated Populations

This paper reports the results of the deployment of Rich-State Simulated Populations at Meta for both automated and manual testing. We use simulated users (aka test users) to mimic user interactions and acquire state in much the same way that real user accounts acquire state. For automated testing, we present empirical results from deployment on the Facebook, Messenger, and Instagram apps for iOS and Android Platforms. These apps consist of tens of millions of lines of code, communicating with hundreds of millions of lines of backend code, and are used by over 2 billion people every day. Our results reveal that rich state increases average code coverage by 38\%, and endpoint coverage by 61\%. More importantly, it also yields an average increase of 115\% in the faults found by automated testing. The rich-state test user populations are also deployed in a (continually evolving) Test Universe; a web-enabled simulation platform for privacy-safe manual testing, which has been used by over 21,000 Meta engineers since its deployment in November 2022.

cs.SE↗

arXiv2024

Fairness Testing: A Comprehensive Survey and Analysis of Trends

Unfair behaviors of Machine Learning (ML) software have garnered increasing attention and concern among software engineers. To tackle this issue, extensive research has been dedicated to conducting fairness testing of ML software, and this paper offers a comprehensive survey of existing studies in this field. We collect 100 papers and organize them based on the testing workflow (i.e., how to test) and testing components (i.e., what to test). Furthermore, we analyze the research focus, trends, and promising directions in the realm of fairness testing. We also identify widely-adopted datasets and open-source tools for fairness testing.

cs.SE↗

arXiv2024

Automated Unit Test Improvement using Large Language Models at Meta

This paper describes Meta's TestGen-LLM tool, which uses LLMs to automatically improve existing human-written tests. TestGen-LLM verifies that its generated test classes successfully clear a set of filters that assure measurable improvement over the original test suite, thereby eliminating problems due to LLM hallucination. We describe the deployment of TestGen-LLM at Meta test-a-thons for the Instagram and Facebook platforms. In an evaluation on Reels and Stories products for Instagram, 75% of TestGen-LLM's test cases built correctly, 57% passed reliably, and 25% increased coverage. During Meta's Instagram and Facebook test-a-thons, it improved 11.5% of all classes to which it was applied, with 73% of its recommendations being accepted for production deployment by Meta software engineers. We believe this is the first report on industrial scale deployment of LLM-generated code backed by such assurances of code improvement.

cs.SE↗

arXiv2024

Observation-based unit test generation at Meta

TestGen automatically generates unit tests, carved from serialized observations of complex objects, observed during app execution. We describe the development and deployment of TestGen at Meta. In particular, we focus on the scalability challenges overcome during development in order to deploy observation-based test carving at scale in industry. So far, TestGen has landed 518 tests into production, which have been executed 9,617,349 times in continuous integration, finding 5,702 faults. Meta is currently in the process of more widespread deployment. Our evaluation reveals that, when carving its observations from 4,361 reliable end-to-end tests, TestGen was able to generate tests for at least 86\% of the classes covered by end-to-end tests. Testing on 16 Kotlin Instagram app-launch-blocking tasks demonstrated that the TestGen tests would have trapped 13 of these before they became launch blocking.

cs.SE↗

arXiv2024

Assured LLM-Based Software Engineering

In this paper we address the following question: How can we use Large Language Models (LLMs) to improve code independently of a human, while ensuring that the improved code - does not regress the properties of the original code? - improves the original in a verifiable and measurable way? To address this question, we advocate Assured LLM-Based Software Engineering; a generate-and-test approach, inspired by Genetic Improvement. Assured LLMSE applies a series of semantic filters that discard code that fails to meet these twin guarantees. This overcomes the potential problem of LLM's propensity to hallucinate. It allows us to generate code using LLMs, independently of any human. The human plays the role only of final code reviewer, as they would do with code generated by other human engineers. This paper is an outline of the content of the keynote by Mark Harman at the International Workshop on Interpretability, Robustness, and Benchmarking in Neural Software Engineering, Monday 15th April 2024, Lisbon, Portugal.

cs.SE↗

arXiv2023

Large Language Models for Software Engineering: Survey and Open Problems

This paper provides a survey of the emerging area of Large Language Models (LLMs) for Software Engineering (SE). It also sets out open research challenges for the application of LLMs to technical problems faced by software engineers. LLMs' emergent properties bring novelty and creativity with applications right across the spectrum of Software Engineering activities including coding, design, requirements, repair, refactoring, performance improvement, documentation and analytics. However, these very same emergent properties also pose significant technical challenges; we need techniques that can reliably weed out incorrect solutions, such as hallucinations. Our survey reveals the pivotal role that hybrid techniques (traditional SE plus LLMs) have to play in the development and deployment of reliable, efficient and effective LLM-based SE.

cs.SE↗

arXiv2023

Bias Mitigation for Machine Learning Classifiers: A Comprehensive Survey

This paper 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.

cs.LG↗

arXiv2023

Large Language Models in Fault Localisation

Large Language Models (LLMs) have shown promise in multiple software engineering tasks including code generation, program repair, code summarisation, and test generation. Fault localisation is instrumental in enabling automated debugging and repair of programs and was prominently featured as a highlight during the launch event of ChatGPT-4. Nevertheless, the performance of LLMs compared to state-of-the-art methods, as well as the impact of prompt design and context length on their efficacy, remains unclear. To fill this gap, this paper presents an in-depth investigation into the capability of ChatGPT-3.5 and ChatGPT-4, the two state-of-the-art LLMs, on fault localisation. Using the widely-adopted large-scale Defects4J dataset, we compare the two LLMs with the existing fault localisation techniques. We also investigate the consistency of LLMs in fault localisation, as well as how prompt engineering and the length of code context affect the fault localisation effectiveness. Our findings demonstrate that within function-level context, ChatGPT-4 outperforms all the existing fault localisation methods. Additional error logs can further improve ChatGPT models' localisation accuracy and consistency, with an average 46.9% higher accuracy over the state-of-the-art baseline SmartFL on the Defects4J dataset in terms of TOP-1 metric. However, when the code context of the Defects4J dataset expands to the class-level, ChatGPT-4's performance suffers a significant drop, with 49.9% lower accuracy than SmartFL under TOP-1 metric. These observations indicate that although ChatGPT can effectively localise faults under specific conditions, limitations are evident. Further research is needed to fully harness the potential of LLMs like ChatGPT for practical fault localisation applications.

cs.SE↗

arXiv2023

COCO: Testing Code Generation Systems via Concretized Instructions

Code generation systems have been extensively developed in recent years to generate source code based on natural language instructions. However, despite their advancements, these systems still face robustness issues where even slightly different instructions can result in significantly different code semantics. Robustness is critical for code generation systems, as it can have significant impacts on software development, software quality, and trust in the generated code. Although existing testing techniques for general text-to-text software can detect some robustness issues, they are limited in effectiveness due to ignoring the characteristics of code generation systems. In this work, we propose a novel technique COCO to test the robustness of code generation systems. It exploits the usage scenario of code generation systems to make the original programming instruction more concrete by incorporating features known to be contained in the original code. A robust system should maintain code semantics for the concretized instruction, and COCO detects robustness inconsistencies when it does not. We evaluated COCO on eight advanced code generation systems, including commercial tools such as Copilot and ChatGPT, using two widely-used datasets. Our results demonstrate the effectiveness of COCO in testing the robustness of code generation systems, outperforming two techniques adopted from general text-to-text software testing by 466.66% and 104.02%, respectively. Furthermore, concretized instructions generated by COCO can help reduce robustness inconsistencies by 18.35% to 53.91% through fine-tuning.

cs.SE↗

arXiv2023

A Comprehensive Empirical Study of Bias Mitigation Methods for Machine Learning Classifiers

Software bias is an increasingly important operational concern for software engineers. We present a large-scale, comprehensive empirical study of 17 representative bias mitigation methods for Machine Learning (ML) classifiers, evaluated with 11 ML performance metrics (e.g., accuracy), 4 fairness metrics, and 20 types of fairness-performance trade-off assessment, applied to 8 widely-adopted software decision tasks. The empirical coverage is much more comprehensive, covering the largest numbers of bias mitigation methods, evaluation metrics, and fairness-performance trade-off measures compared to previous work on this important software property. We find that (1) the bias mitigation methods significantly decrease ML performance in 53% of the studied scenarios (ranging between 42%~66% according to different ML performance metrics); (2) the bias mitigation methods significantly improve fairness measured by the 4 used metrics in 46% of all the scenarios (ranging between 24%~59% according to different fairness metrics); (3) the bias mitigation methods even lead to decrease in both fairness and ML performance in 25% of the scenarios; (4) the effectiveness of the bias mitigation methods depends on tasks, models, the choice of protected attributes, and the set of metrics used to assess fairness and ML performance; (5) there is no bias mitigation method that can achieve the best trade-off in all the scenarios. The best method that we find outperforms other methods in 30% of the scenarios. Researchers and practitioners need to choose the bias mitigation method best suited to their intended application scenario(s).

cs.SE↗

arXiv2023

Simulation-Driven Automated End-to-End Test and Oracle Inference

This is the first work to report on inferential testing at scale in industry. Specifically, it reports the experience of automated testing of integrity systems at Meta. We built an internal tool called ALPACAS for automated inference of end-to-end integrity tests. Integrity tests are designed to keep users safe online by checking that interventions take place when harmful behaviour occurs on a platform. ALPACAS infers not only the test input, but also the oracle, by observing production interventions to prevent harmful behaviour. This approach allows Meta to automate the process of generating integrity tests for its platforms, such as Facebook and Instagram, which consist of hundreds of millions of lines of production code. We outline the design and deployment of ALPACAS, and report results for its coverage, number of tests produced at each stage of the test inference process, and their pass rates. Specifically, we demonstrate that using ALPACAS significantly improves coverage from a manual test design for the particular aspect of integrity end-to-end testing it was applied to. Further, from a pool of 3 million data points, ALPACAS automatically yields 39 production-ready end-to-end integrity tests. We also report that the ALPACAS-inferred test suite enjoys exceptionally low flakiness for end-to-end testing with its average in-production pass rate of 99.84%.

cs.SE↗

arXiv2022

Keeping Mutation Test Suites Consistent and Relevant with Long-Standing Mutants

Mutation testing has been demonstrated to be one of the most powerful fault-revealing tools in the tester's tool kit. Much previous work implicitly assumed it to be sufficient to re-compute mutant suites per release. Sadly, this makes mutation results inconsistent; mutant scores from each release cannot be directly compared, making it harder to measure test improvement. Furthermore, regular code change means that a mutant suite's relevance will naturally degrade over time. We measure this degradation in relevance for 143,500 mutants in 4 non-trivial systems finding that, on overage, 52% degrade. We introduce a mutant brittleness measure and use it to audit software systems and their mutation suites. We also demonstrate how consistent-by-construction long-standing mutant suites can be identified with a 10x improvement in mutant relevance over an arbitrary test suite. Our results indicate that the research community should avoid the re-computation of mutant suites and focus, instead, on long-standing mutants, thereby improving the consistency and relevance of mutation testing.

cs.SE↗

arXiv2022

Leveraging Automated Unit Tests for Unsupervised Code Translation

With little to no parallel data available for programming languages, unsupervised methods are well-suited to source code translation. However, the majority of unsupervised machine translation approaches rely on back-translation, a method developed in the context of natural language translation and one that inherently involves training on noisy inputs. Unfortunately, source code is highly sensitive to small changes; a single token can result in compilation failures or erroneous programs, unlike natural languages where small inaccuracies may not change the meaning of a sentence. To address this issue, we propose to leverage an automated unit-testing system to filter out invalid translations, thereby creating a fully tested parallel corpus. We found that fine-tuning an unsupervised model with this filtered data set significantly reduces the noise in the translations so-generated, comfortably outperforming the state-of-the-art for all language pairs studied. In particular, for Java $\to$ Python and Python $\to$ C++ we outperform the best previous methods by more than 16% and 24% respectively, reducing the error rate by more than 35%.

cs.SE↗

arXiv2021

Model Validation Using Mutated Training Labels: An Exploratory Study

We introduce an exploratory study on Mutation Validation (MV), a model validation method using mutated training labels for supervised learning. MV mutates training data labels, retrains the model against the mutated data, then uses the metamorphic relation that captures the consequent training performance changes to assess model fit. It does not use a validation set or test set. The intuition underpinning MV is that overfitting models tend to fit noise in the training data. We explore 8 different learning algorithms, 18 datasets, and 5 types of hyperparameter tuning tasks. Our results demonstrate that MV is accurate in model selection: the model recommendation hit rate is 92\% for MV and less than 60\% for out-of-sample-validation. MV also provides more stable hyperparameter tuning results than out-of-sample-validation across different runs.

cs.LG↗