Shin Yoo

Juyeon Yoon, Robert Feldt, Shin Yoo

GUI testing checks if a software system behaves as expected when users interact with its graphical interface, e.g., testing specific functionality or validating relevant use case scenarios. Currently, deciding what to test at this high level is a manual task since automated GUI testing tools target lower level adequacy metrics such as structural code coverage or activity coverage. We propose DroidAgent, an autonomous GUI testing agent for Android, for semantic, intent-driven automation of GUI testing. It is based on Large Language Models and support mechanisms such as long- and short-term memory. Given an Android app, DroidAgent sets relevant task goals and subsequently tries to achieve them by interacting with the app. Our empirical evaluation of DroidAgent using 15 apps from the Themis benchmark shows that it can set up and perform realistic tasks, with a higher level of autonomy. For example, when testing a messaging app, DroidAgent created a second account and added a first account as a friend, testing a realistic use case, without human intervention. On average, DroidAgent achieved 61% activity coverage, compared to 51% for current state-of-the-art GUI testing techniques. Further, manual analysis shows that 317 out of the 374 autonomously created tasks are realistic and relevant to app functionalities, and also that DroidAgent interacts deeply with the apps and covers more features.

Naryeong Kim, Shin Yoo

Open-weight Small Language Models(SLMs) can provide faster local inference at lower financial cost, but may not achieve the same performance level as commercial Large Language Models (LLMs) that are orders of magnitudes larger. Consequently, many of the latest applications of LLMs, such as software engineering agents, tend to be evaluated on larger models only, leaving the issue of improving the cost-benefit trade-off of such applications neglected. This paper proposes Atropos, a predictive early-termination analysis and hotswap technique that aims to improve the cost-benefit trade-off for LLM-based agents that use self-consistency. The core component of ATROPOS is a predictive model based on structural properties of LLM inferences: after merging multiple agentic inference paths into a graph representation, ATROPOS uses Graph Convolutional Network (GCN) to predict whether an ongoing inference will eventually succeed or not. If an agentic task instance running on the source LLM is predicted to fail, ATROPOS subsequently performs hotswapping, i.e., migrating the on-going inference context onto the more capable target LLM: this is feasible because LLM contexts are stateless. An empirical evaluation of ATROPOS using three recent LLM-based agents shows that ATROPOS can predict early termination of eventually failing inferences with the accuracy of 0.85 at the midpoint of the inference. Hotswapping LLMs for such inferences can convert up to 27.57% of them to be successful. Consequently, ATROPOS achieves 74.35% of the performance of closed LLMs with as low as only 23.9% of the cost.

Hyunjoon Cho, Sungmin Kang, Gabin An et al.

LLMs are rapidly being adopted to build powerful tools and agents for software engineering, but most of them rely heavily on extremely large closed-source models. This, in turn, can hinder wider adoption due to security issues as well as financial cost and environmental impact. Recently, a number of open source Small Language Models (SLMs) are being released and gaining traction. While SLMs are smaller, more energy-efficient, and therefore easier to locally deploy, they tend to show worse performance when compared to larger closed LLMs. We present COSMos, a task-level LLM ensemble technique that uses voting mechanism, to provide a broader range of choice between SLMs and LLMs. We instantiate COSMos with an LLM-based Fault Localisation technique, AutoFL, and report the cost-benefit trade-off between LLM accuracy and various costs such as energy consumption, inference time, and the number of tokens used. An empirical evaluation using Defects4J shows that COSMos can build effective ensembles that can achieve Pareto-optimality in terms of FL accuracy and inference cost, when compared to individual models.

Junho Myung, Yeon Su Park, Sunwoo Kim et al.

Evaluating the performance and biases of large language models (LLMs) through role-playing scenarios is becoming increasingly common, as LLMs often exhibit biased behaviors in these contexts. Building on this line of research, we introduce PapersPlease, a benchmark consisting of 3,700 moral dilemmas designed to investigate LLMs' decision-making in prioritizing various levels of human needs. In our setup, LLMs act as immigration inspectors deciding whether to approve or deny entry based on the short narratives of people. These narratives are constructed using the Existence, Relatedness, and Growth (ERG) theory, which categorizes human needs into three hierarchical levels. Our analysis of six LLMs reveals statistically significant patterns in decision-making, suggesting that LLMs encode implicit preferences. Additionally, our evaluation of the impact of incorporating social identities into the narratives shows varying responsiveness based on both motivational needs and identity cues, with some models exhibiting higher denial rates for marginalized identities. All data is publicly available at https://github.com/yeonsuuuu28/papers-please.

Felix Dobslaw, Robert Feldt, Juyeon Yoon et al.

Large Language Models (LLMs) and Multi-Agent LLMs (MALLMs) introduce non-determinism unlike traditional or machine learning software, requiring new approaches to verifying correctness beyond simple output comparisons or statistical accuracy over test datasets. This paper presents a taxonomy for LLM test case design, informed by research literature and our experience. Each facet is exemplified, and we conduct an LLM-assisted analysis of six open-source testing frameworks, perform a sensitivity study of an agent-based system across different model configurations, and provide working examples contrasting atomic and aggregated test cases. We identify key variation points that impact test correctness and highlight open challenges that the research, industry, and open-source communities must address as LLMs become integral to software systems. Our taxonomy defines four facets of LLM test case design, addressing ambiguity in both inputs and outputs while establishing best practices. It distinguishes variability in goals, the system under test, and inputs, and introduces two key oracle types: atomic and aggregated. Our findings reveal that current tools treat test executions as isolated events, lack explicit aggregation mechanisms, and inadequately capture variability across model versions, configurations, and repeated runs. This highlights the need for viewing correctness as a distribution of outcomes rather than a binary property, requiring closer collaboration between academia and practitioners to establish mature, variability-aware testing methodologies.

Saeyoon Oh, Shin Yoo

When applying the Transformer architecture to source code, designing a good self-attention mechanism is critical as it affects how node relationship is extracted from the Abstract Syntax Trees (ASTs) of the source code. We present Code Structure Aware Transformer (CSA-Trans), which uses Code Structure Embedder (CSE) to generate specific PE for each node in AST. CSE generates node Positional Encoding (PE) using disentangled attention. To further extend the self-attention capability, we adopt Stochastic Block Model (SBM) attention. Our evaluation shows that our PE captures the relationships between AST nodes better than other graph-related PE techniques. We also show through quantitative and qualitative analysis that SBM attention is able to generate more node specific attention coefficients. We demonstrate that CSA-Trans outperforms 14 baselines in code summarization tasks for both Python and Java, while being 41.92% faster and 25.31% memory efficient in Java dataset compared to AST-Trans and SG-Trans respectively.

Juyeon Yoon, Robert Feldt, Shin Yoo

The recent surge of building software systems powered by Large Language Models (LLMs) has led to the development of various testing frameworks, primarily focused on treating prompt templates as the unit of testing. Despite the significant costs associated with test input execution and output assessment, the curation of optimized test suites is yet overlooked in these tools, which calls for tailored test selection or prioritization strategies. In this paper, we show that diversity-based testing techniques, such as Adaptive Random Testing (ART) with appropriate string distance metrics, can be effectively applied to the testing of prompt templates. Our proposed adaptive testing approach adjusts the conventional ART process to this context by selecting new test inputs based on scores derived from existing test suite and their labelling results. Our results, obtained using various implementations that explore several string-based distances, confirm that our approach enables the discovery of failures with reduced testing budgets and promotes the generation of more varied outputs.

Jae Yong Lee, Sungmin Kang, Shin Yoo

Large Language Models (LLMs) are machine learning models that have seen widespread adoption due to their capability of handling previously difficult tasks. LLMs, due to their training, are sensitive to how exactly a question is presented, also known as prompting. However, prompting well is challenging, as it has been difficult to uncover principles behind prompting -- generally, trial-and-error is the most common way of improving prompts, despite its significant computational cost. In this context, we argue it would be useful to perform `predictive prompt analysis', in which an automated technique would perform a quick analysis of a prompt and predict how the LLM would react to it, relative to a goal provided by the user. As a demonstration of the concept, we present Syntactic Prevalence Analyzer (SPA), a predictive prompt analysis approach based on sparse autoencoders (SAEs). SPA accurately predicted how often an LLM would generate target syntactic structures during code synthesis, with up to 0.994 Pearson correlation between the predicted and actual prevalence of the target structure. At the same time, SPA requires only 0.4\% of the time it takes to run the LLM on a benchmark. As LLMs are increasingly used during and integrated into modern software development, our proposed predictive prompt analysis concept has the potential to significantly ease the use of LLMs for both practitioners and researchers.

Gunel Jahangirova, Nargiz Humbatova, Jinhan Kim et al.

As the adoption of Deep Learning (DL) systems continues to rise, an increasing number of approaches are being proposed to test these systems, localise faults within them, and repair those faults. The best attestation of effectiveness for such techniques is an evaluation that showcases their capability to detect, localise and fix real faults. To facilitate these evaluations, the research community has collected multiple benchmarks of real faults in DL systems. In this work, we perform a manual analysis of 490 faults from five different benchmarks and identify that 314 of them are eligible for our study. Our investigation focuses specifically on how well the bugs correspond to the sources they were extracted from, which fault types are represented, and whether the bugs are reproducible. Our findings indicate that only 18.5% of the faults satisfy our realism conditions. Our attempts to reproduce these faults were successful only in 52% of cases.

Nargiz Humbatova, Jinhan Kim, Gunel Jahangirova et al.

With the increased popularity of Deep Neural Networks (DNNs), increases also the need for tools to assist developers in the DNN implementation, testing and debugging process. Several approaches have been proposed that automatically analyse and localise potential faults in DNNs under test. In this work, we evaluate and compare existing state-of-the-art fault localisation techniques, which operate based on both dynamic and static analysis of the DNN. The evaluation is performed on a benchmark consisting of both real faults obtained from bug reporting platforms and faulty models produced by a mutation tool. Our findings indicate that the usage of a single, specific ground truth (e.g., the human defined one) for the evaluation of DNN fault localisation tools results in pretty low performance (maximum average recall of 0.31 and precision of 0.23). However, such figures increase when considering alternative, equivalent patches that exist for a given faulty DNN. Results indicate that \dfd is the most effective tool, achieving an average recall of 0.61 and precision of 0.41 on our benchmark.

Juyeon Yoon, Somin Kim, Robert Feldt et al.

Software increasingly relies on the emergent capabilities of Large Language Models (LLMs), from natural language understanding to program analysis and generation. Yet testing them on specific tasks remains difficult and costly: many prompts lack ground truth, forcing reliance on human judgment, while existing uncertainty and adequacy measures typically require full inference. A key challenge is to assess input adequacy in a way that reflects the demands of the task, ideally before even generating any output. We introduce CLOTHO, a task-specific, pre-generation adequacy measure that estimates input difficulty directly from hidden LLM states. Given a large pool of unlabelled inputs for a specific task, CLOTHO uses a Gaussian Mixture Model (GMM) to adaptively sample the most informative cases for human labelling. Based on this reference set the GMM can then rank unseen inputs by their likelihood of failure. In our empirical evaluation across eight benchmark tasks and three open-weight LLMs, CLOTHO can predict failures with a ROC-AUC of 0.716, after labelling reference sets that are on average only 5.4% of inputs. It does so without generating any outputs, thereby reducing costs compared to existing uncertainty measures. Comparison of CLOTHO and post-generation uncertainty measures shows that the two approaches complement each other. Crucially, we show that adequacy scores learnt from open-weight LLMs transfer effectively to proprietary models, extending the applicability of the approach. When prioritising test inputs for proprietary models, CLOTHO increases the average number of failing inputs from 18.7 to 42.5 out of 100, compared to random prioritisation.

Shin Yoo, Robert Feldt, Somin Kim et al.

ML-based systems are software systems that incorporates machine learning components such as Deep Neural Networks (DNNs) or Large Language Models (LLMs). While such systems enable advanced features such as high performance computer vision, natural language processing, and code generation, their internal behaviour remain largely opaque to traditional dynamic analysis such as testing: existing analysis typically concern only what is observable from the outside, such as input similarity or class label changes. We propose semantic flow, a concept designed to capture the internal behaviour of ML-based system and to provide a platform for traditional dynamic analysis techniques to be adapted to. Semantic flow combines the idea of control flow with internal states taken from executions of ML-based systems, such as activation values of a specific layer in a DNN, or embeddings of LLM responses at a specific inference step of LLM agents. The resulting representation, summarised as semantic flow graphs, can capture internal decisions that are not explicitly represented in the traditional control flow of ML-based systems. We propose the idea of semantic flow, introduce two examples using a DNN and an LLM agent, and finally sketch its properties and how it can be used to adapt existing dynamic analysis techniques for use in ML-based software systems.

Somin Kim, Shin Yoo

Surprise Adequacy (SA) has been widely studied as a test adequacy metric that can effectively guide software engineers towards inputs that are more likely to reveal unexpected behaviour of Deep Neural Networks (DNNs). Intuitively, SA is an out-of-distribution metric that quantifies the dissimilarity between the given input and the training data: if a new input is very different from those seen during training, the DNN is more likely to behave unexpectedly against the input. While SA has been widely adopted as a test prioritization method, its major weakness is the fact that the computation of the metric requires access to the training dataset, which is often not allowed in real-world use cases. We present DANDI, a technique that generates a surrogate input distribution using Stable Diffusion to compute SA values without requiring the original training data. An empirical evaluation of DANDI applied to image classifiers for CIFAR10 and ImageNet-1K shows that SA values computed against synthetic data are highly correlated with the values computed against the training data, with Spearman Rank correlation value of 0.852 for ImageNet-1K and 0.881 for CIFAR-10. Further, we show that SA value computed by DANDI achieves can prioritize inputs as effectively as those computed using the training data, when testing DNN models mutated by DeepMutation. We believe that DANDI can significantly improve the usability of SA for practical DNN testing.

Gabin An, Juyeon Yoon, Shin Yoo

Defects4J has enabled numerous software testing and debugging research work since its introduction. A large part of its contribution, and the resulting popularity, lies in the clear separation and distillation of the root cause of each individual test failure based on careful manual analysis, which in turn allowed researchers to easily study individual faults in isolation. However, in a realistic debugging scenario, multiple faults can coexist and affect test results collectively. Study of automated debugging techniques for these situations, such as failure clustering or fault localisation for multiple faults, would significantly benefit from a reliable benchmark of multiple, coexisting faults. We search for versions of Defects4J subjects that contain multiple faults, by iteratively transplanting fault-revealing test cases across Defects4J versions. Out of 326 studied versions of Defects4J subjects, we report that over 95% (311 versions) actually contain from two to 24 faults. We hope that the extended, multi-fault Defects4J can provide a platform for future research of testing and debugging techniques for multi-fault programs.

Saeyoon Oh, Seongmin Lee, Shin Yoo

Higher Order Mutation (HOM) has been proposed to avoid equivalent mutants and improve the scalability of mutation testing, but generating useful HOMs remain an expensive search problem on its own. We propose a new approach to generate Strongly Subsuming Higher Order Mutants (SSHOM) using a recently introduced Causal Program Dependence Analysis (CPDA). CPDA itself is based on program mutation, and provides quantitative estimation of how often a change of the value of a program element will cause a value change of another program element. Our SSHOM generation approach chooses pairs of program elements using heuristics based on CPDA analysis, performs First Order Mutation to the chosen pairs, and generates an HOM by combining two FOMs.

Jinhan Kim, Juyoung Jeon, Shin Hong et al.

Mutation analysis can provide valuable insights into both System Under Test (SUT) and its test suite. However, it is not scalable due to the cost of building and testing a large number of mutants. Predictive Mutation Testing (PMT) has been proposed to reduce the cost of mutation testing, but it can only provide statistical inference about whether a mutant will be killed or not by the entire test suite. We propose Seshat, a Predictive Mutation Analysis (PMA) technique that can accurately predict the entire kill matrix, not just the mutation score of the given test suite. Seshat exploits the natural language channel in code, and learns the relationship between the syntactic and semantic concepts of each test case and the mutants it can kill, from a given kill matrix. The learnt model can later be used to predict the kill matrices for subsequent versions of the program, even after both the source and test code have changed significantly. Empirical evaluation using the programs in the Defects4J shows that Seshat can predict kill matrices with the average F-score of 0.83 for versions that are up to years apart. This is an improvement of F-score by 0.14 and 0.45 point over the state-of-the-art predictive mutation testing technique, and a simple coverage based heuristic, respectively. Seshat also performs as well as PMT for the prediction of mutation scores only. Once Seshat trains its model using a concrete mutation analysis, the subsequent predictions made by Seshat are on average 39 times faster than actual test-based analysis.

Gabin An, Juyeon Yoon, Joyce Jiyoung Whang et al.

Automated debugging techniques, such as Fault Localisation (FL) or Automated Program Repair (APR), are typically designed under the Single Fault Assumption (SFA). However, in practice, an unknown number of faults can independently cause multiple test case failures, making it difficult to allocate resources for debugging and to use automated debugging techniques. Clustering algorithms have been applied to group the test failures according to their root causes, but their accuracy can often be lacking due to the inherent limits in the distance metrics for test cases. We introduce a new test distance metric based on hypergraphs and evaluate their accuracy using multi-fault benchmarks that we have built on top of Defects4J and SIR. Results show that our technique, Hybiscus, can automatically achieve perfect clustering (i.e., the same number of clusters as the ground truth number of root causes, with all failing tests with the same root cause grouped together) for 418 out of 605 test runs with multiple test failures. Better failure clustering also allows us to separate different root causes and apply FL techniques under SFA, resulting in saving up to 82% of the total wasted effort when compared to the state-of-the-art technique for multiple fault localisation.

Seongmin Lee, Dave Binkley, Robert Feldt et al.

We introduce Causal Program Dependence Analysis (CPDA), a dynamic dependence analysis that applies causal inference to model the strength of program dependence relations in a continuous space. CPDA observes the association between program elements by constructing and executing modified versions of a program. One advantage of CPDA is that this construction requires only light-weight parsing rather than sophisticated static analysis. The result is a collection of observations based on how often a change in the value produced by a mutated program element affects the behavior of other elements. From this set of observations, CPDA discovers a causal structure capturing the causal (i.e., dependence) relation between program elements. Qualitative evaluation finds that CPDA concisely expresses key dependence relationships between program elements. As an example application, we apply CPDA to the problem of fault localization. Using minimal test suites, our approach can rank twice as many faults compared to SBFL.

Gabin An, Shin Yoo

The performance of many Fault Localisation (FL) techniques directly depends on the quality of the used test suites. Consequently, it is extremely useful to be able to precisely measure how much diagnostic power each test case can introduce when added to a test suite used for FL. Such a measure can help us not only to prioritise and select test cases to be used for FL, but also to effectively augment test suites that are too weak to be used with FL techniques. We propose FDG, a new measure of Fault Diagnosability Gain for individual test cases. The design of FDG is based on our analysis of existing metrics that are designed to prioritise test cases for better FL. Unlike other metrics, FDG exploits the ongoing FL results to emphasise the parts of the program for which more information is needed. Our evaluation of FDG with Defects4J shows that it can successfully help the augmentation of test suites for better FL. When given only a few failing test cases (2.3 test cases on average), FDG can effectively augment the given test suite by prioritising the test cases generated automatically by EvoSuite: the augmentation can improve the acc@1 and acc@10 of the FL results by 11.6x and 2.2x on average, after requiring only ten human judgements on the correctness of the assertions EvoSuite generates.

William B. Langdon, Westley Weimer, Justyna Petke et al.

Following Prof. Mark Harman of Facebook's keynote and formal presentations (which are recorded in the proceedings) there was a wide ranging discussion at the eighth international Genetic Improvement workshop, GI-2020 @ ICSE (held as part of the 42nd ACM/IEEE International Conference on Software Engineering on Friday 3rd July 2020). Topics included industry take up, human factors, explainabiloity (explainability, justifyability, exploitability) and GI benchmarks. We also contrast various recent online approaches (e.g. SBST 2020) to holding virtual computer science conferences and workshops via the WWW on the Internet without face-2-face interaction. Finally we speculate on how the Coronavirus Covid-19 Pandemic will affect research next year and into the future.

Jinhan Kim, Jeongil Ju, Robert Feldt et al.

Deep Neural Networks (DNNs) are rapidly being adopted by the automotive industry, due to their impressive performance in tasks that are essential for autonomous driving. Object segmentation is one such task: its aim is to precisely locate boundaries of objects and classify the identified objects, helping autonomous cars to recognise the road environment and the traffic situation. Not only is this task safety critical, but developing a DNN based object segmentation module presents a set of challenges that are significantly different from traditional development of safety critical software. The development process in use consists of multiple iterations of data collection, labelling, training, and evaluation. Among these stages, training and evaluation are computation intensive while data collection and labelling are manual labour intensive. This paper shows how development of DNN based object segmentation can be improved by exploiting the correlation between Surprise Adequacy (SA) and model performance. The correlation allows us to predict model performance for inputs without manually labelling them. This, in turn, enables understanding of model performance, more guided data collection, and informed decisions about further training. In our industrial case study the technique allows cost savings of up to 50% with negligible evaluation inaccuracy. Furthermore, engineers can trade off cost savings versus the tolerable level of inaccuracy depending on different development phases and scenarios.

Sungmin Kang, Robert Feldt, Shin Yoo

The testing of Deep Neural Networks (DNNs) has become increasingly important as DNNs are widely adopted by safety critical systems. While many test adequacy criteria have been suggested, automated test input generation for many types of DNNs remains a challenge because the raw input space is too large to randomly sample or to navigate and search for plausible inputs. Consequently, current testing techniques for DNNs depend on small local perturbations to existing inputs, based on the metamorphic testing principle. We propose new ways to search not over the entire image space, but rather over a plausible input space that resembles the true training distribution. This space is constructed using Variational Autoencoders (VAEs), and navigated through their latent vector space. We show that this space helps efficiently produce test inputs that can reveal information about the robustness of DNNs when dealing with realistic tests, opening the field to meaningful exploration through the space of highly structured images.

Paul Ralph, Sebastian Baltes, Gianisa Adisaputri et al.

Context. As a novel coronavirus swept the world in early 2020, thousands of software developers began working from home. Many did so on short notice, under difficult and stressful conditions. Objective. This study investigates the effects of the pandemic on developers' wellbeing and productivity. Method. A questionnaire survey was created mainly from existing, validated scales and translated into 12 languages. The data was analyzed using non-parametric inferential statistics and structural equation modeling. Results. The questionnaire received 2225 usable responses from 53 countries. Factor analysis supported the validity of the scales and the structural model achieved a good fit (CFI = 0.961, RMSEA = 0.051, SRMR = 0.067). Confirmatory results include: (1) the pandemic has had a negative effect on developers' wellbeing and productivity; (2) productivity and wellbeing are closely related; (3) disaster preparedness, fear related to the pandemic and home office ergonomics all affect wellbeing or productivity. Exploratory analysis suggests that: (1) women, parents and people with disabilities may be disproportionately affected; (2) different people need different kinds of support. Conclusions. To improve employee productivity, software companies should focus on maximizing employee wellbeing and improving the ergonomics of employees' home offices. Women, parents and disabled persons may require extra support.

Jeongju Sohn, Sungmin Kang, Shin Yoo

The rapid and widespread adoption of Deep Neural Networks (DNNs) has called for ways to test their behaviour, and many testing approaches have successfully revealed misbehaviour of DNNs. However, it is relatively unclear what one can do to correct such behaviour after revelation, as retraining involves costly data collection and does not guarantee to fix the underlying issue. This paper introduces Arachne, a novel program repair technique for DNNs, which directly repairs DNNs using their input-output pairs as a specification. Arachne localises neural weights on which it can generate effective patches and uses Differential Evolution to optimise the localised weights and correct the misbehaviour. An empirical study using different benchmarks shows that Arachne can fix specific misclassifications of a DNN without reducing general accuracy significantly. On average, patches generated by Arachne generalise to 61.3% of unseen misbehaviour, whereas those by a state-of-the-art DNN repair technique generalise only to 10.2% and sometimes to none while taking tens of times more than Arachne. We also show that Arachne can address fairness issues by debiasing a gender classification model. Finally, we successfully apply Arachne to a text sentiment model to show that it generalises beyond Convolutional Neural Networks.

Jinhan Kim, Gabin An, Robert Feldt et al.

Mutation analysis can effectively capture the dependency between source code and test results. This has been exploited by Mutation Based Fault Localisation (MBFL) techniques. However, MBFL techniques suffer from the need to expend the high cost of mutation analysis after the observation of failures, which may present a challenge for its practical adoption. We introduce SIMFL (Statistical Inference for Mutation-based Fault Localisation), an MBFL technique that allows users to perform the mutation analysis in advance before a failure is observed, allowing the amortisation of the analysis cost. SIMFL uses mutants as artificial faults and aims to learn the failure patterns among test cases against different locations of mutations. Once a failure is observed, SIMFL requires either almost no or very small additional cost for analysis, depending on the used inference model. An empirical evaluation using Defects4J shows that SIMFL can successfully localise up to 113 out of 203 studied faults (55%) at the top, and 159 (78%) faults within the top five, significantly outperforming existing MBFL techniques while using the results of mutation analysis that has been undertaken before the test failure. The amortised cost of mutation analysis can be further reduced by mutation sampling: SIMFL retains 80% of its localisation accuracy at the top rank when using only 10% of generated mutants, compared to results obtained without sampling.

Jinhan Kim, Robert Feldt, Shin Yoo

Deep Learning (DL) systems are rapidly being adopted in safety and security critical domains, urgently calling for ways to test their correctness and robustness. Testing of DL systems has traditionally relied on manual collection and labelling of data. Recently, a number of coverage criteria based on neuron activation values have been proposed. These criteria essentially count the number of neurons whose activation during the execution of a DL system satisfied certain properties, such as being above predefined thresholds. However, existing coverage criteria are not sufficiently fine grained to capture subtle behaviours exhibited by DL systems. Moreover, evaluations have focused on showing correlation between adversarial examples and proposed criteria rather than evaluating and guiding their use for actual testing of DL systems. We propose a novel test adequacy criterion for testing of DL systems, called Surprise Adequacy for Deep Learning Systems (SADL), which is based on the behaviour of DL systems with respect to their training data. We measure the surprise of an input as the difference in DL system's behaviour between the input and the training data (i.e., what was learnt during training), and subsequently develop this as an adequacy criterion: a good test input should be sufficiently but not overtly surprising compared to training data. Empirical evaluation using a range of DL systems from simple image classifiers to autonomous driving car platforms shows that systematic sampling of inputs based on their surprise can improve classification accuracy of DL systems against adversarial examples by up to 77.5% via retraining.

Kui Liu, Dongsun Kim, Tegawendé F. Bissyandé et al.

In this paper, we first collect and track a large number of fixed and unfixed violations across revisions of software. The empirical analyses reveal that there are discrepancies in the distributions of violations that are detected and those that are fixed, in terms of occurrences, spread and categories, which can provide insights into prioritizing violations. To automatically identify patterns in violations and their fixes, we propose an approach that utilizes convolutional neural networks to learn features and clustering to regroup similar instances. We then evaluate the usefulness of the identified fix patterns by applying them to unfixed violations. The results show that developers will accept and merge a majority (69/116) of fixes generated from the inferred fix patterns. It is also noteworthy that the yielded patterns are applicable to four real bugs in the Defects4J major benchmark for software testing and automated repair.

Donghwan Shin, Shin Yoo, Mike Papadakis et al.

We propose a new test case prioritization technique that combines both mutation-based and diversity-based approaches. Our diversity-aware mutation-based technique relies on the notion of mutant distinguishment, which aims to distinguish one mutant's behavior from another, rather than from the original program. We empirically investigate the relative cost and effectiveness of the mutation-based prioritization techniques (i.e., using both the traditional mutant kill and the proposed mutant distinguishment) with 352 real faults and 553,477 developer-written test cases. The empirical evaluation considers both the traditional and the diversity-aware mutation criteria in various settings: single-objective greedy, hybrid, and multi-objective optimization. The results show that there is no single dominant technique across all the studied faults. To this end, \rev{we we show when and the reason why each one of the mutation-based prioritization criteria performs poorly, using a graphical model called Mutant Distinguishment Graph (MDG) that demonstrates the distribution of the fault detecting test cases with respect to mutant kills and distinguishment.

Robert Feldt, Simon Poulding, David Clark et al.

A common and natural intuition among software testers is that test cases need to differ if a software system is to be tested properly and its quality ensured. Consequently, much research has gone into formulating distance measures for how test cases, their inputs and/or their outputs differ. However, common to these proposals is that they are data type specific and/or calculate the diversity only between pairs of test inputs, traces or outputs. We propose a new metric to measure the diversity of sets of tests: the test set diameter (TSDm). It extends our earlier, pairwise test diversity metrics based on recent advances in information theory regarding the calculation of the normalized compression distance (NCD) for multisets. An advantage is that TSDm can be applied regardless of data type and on any test-related information, not only the test inputs. A downside is the increased computational time compared to competing approaches. Our experiments on four different systems show that the test set diameter can help select test sets with higher structural and fault coverage than random selection even when only applied to test inputs. This can enable early test design and selection, prior to even having a software system to test, and complement other types of test automation and analysis. We argue that this quantification of test set diversity creates a number of opportunities to better understand software quality and provides practical ways to increase it.