Precise event spotting (PES) aims to recognize fine-grained events at exact moments and has become a key component of sports analytics. This task is particularly challenging due to rapid succession, motion blur, and subtle visual differences. Consequently, most existing methods rely on domain-specific, end-to-end training with large labeled datasets and often struggle in few-shot conditions due to their dependence on pixel- or pose-based inputs alone. However, obtaining large labeled datasets is practically hard. We propose a Unified Multi-Entity Graph Network (UMEG-Net) for few-shot PES. UMEG-Net integrates human skeletons and sport-specific object keypoints into a unified graph and features an efficient spatio-temporal extraction module based on advanced GCN and multi-scale temporal shift. To further enhance performance, we employ multimodal distillation to transfer knowledge from keypoint-based graphs to visual representations. Our approach achieves robust performance with limited labeled data and significantly outperforms baseline models in few-shot settings, providing a scalable and effective solution for few-shot PES.

Xing Ma, Yangjie Zhou, Wu Sun, Zihan Liu, Jingwen Leng, Yun Lin, Shixuan Sun, Minyi Guo, Jin Song Dong. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Large language models (LLMs) for code editing have achieved remarkable progress, yet recent empirical studies reveal a fundamental disconnect between technical accuracy and developer productivity . Despite their strong benchmark performance, developers complete tasks 19% slower when using AI assistance, with over 68.81% of recommendations disrupting their mental flow. This misalignment stems from the use of static commit snapshots that lack temporal information, causing models to optimize for end results rather than the incremental, context-sensitive steps that align with developers’ natural reasoning process. To bridge this gap, we present EditFlow , which benchmarks and optimizes subsequent code edit recommendation systems through the reconstruction of developer editing flows. EditFlow addresses three key challenges. First, collecting edit-order data that reflects developers’ flow is inherently difficult: manual annotation introduces prohibitive overhead, while development logs capture only single trajectories instead of all plausible editing flows. Second, benchmarking recommendation performance against developers’ ongoing editing flow requires a digital-twin-like simulation that can faithfully simulate the editing process. Third, existing heterogeneous systems vary drastically in scale and architecture, posing challenges for developing a unified optimization strategy that endows all models with mental-flow awareness regardless of design or capability. To overcome these challenges, we propose three tightly coupled components: (1) a prompt auto-tuning mechanism that learns an optimized prompt for inferring the relative order between two edits, (2) a digital twin that replays reconstructed edit sequences to simulate developers’ editing process, and (3) EditFlow , a unified optimization strategy that optimizes the flow continuity of subsequent edit suggestions based on developers’ ongoing flow. Evaluations across diverse benchmarks, including manually annotated commits, real-world industrial code, and open-source repositories, show that EditFlow improves order reconstruction accuracy by 63.81%, reduces flow violations by over 75%, and boosts recommendation precision by 66.99%. A user study with 32 developers further demonstrates 25.11% faster task completion and significantly higher perceived recommendation quality. To the best of our knowledge, EditFlow is the first to evaluate and optimize code edit recommendation systems from the perspective of developers’ mental flow, establishing flow-awareness as a new dimension for advancing human-AI code collaboration.

GUI test migration aims to produce test cases with events and assertions to test specific functionalities of a target app. Existing migration approaches typically focus on the widget-mapping paradigm that maps widgets from source apps to target apps. However, since different apps may implement the same functionality in different ways, direct mapping may result in incomplete or buggy test cases, thus significantly impacting the effectiveness of testing the target functionality and the practical applicability of migration approaches. In this article, we propose a new migration paradigm (i.e., the abstraction-concretization paradigm) that first abstracts the test logic for the target functionality and then utilizes this logic to generate the concrete GUI test case. Furthermore, we introduce MACdroid , the first approach that migrates GUI test cases based on this paradigm. Specifically, we propose an abstraction technique that utilizes source test cases from source apps targeting the same functionality to extract a general test logic for that functionality. Then, we propose a concretization technique that utilizes the general test logic to guide an LLM in generating the corresponding GUI test case (including events and assertions) for the target app. We evaluate MACdroid on two widely used datasets (including 31 apps, 34 functionalities, and 123 test cases). On the FrUITeR dataset, the test cases generated by MACdroid successfully test 64% of the target functionalities, improving the baselines by 191%. On the Lin dataset, MACdroid successfully tests 75% of the target functionalities, outperforming the baselines by 42%. These results underscore the effectiveness of MACdroid in GUI test migration.

Analyzing Fast, Frequent, and Fine-grained ($F^3$) events presents a significant challenge in video analytics and multi-modal LLMs. Current methods struggle to identify events that satisfy all the $F^3$ criteria with high accuracy due to challenges such as motion blur and subtle visual discrepancies. To advance research in video understanding, we introduce $F^3Set$, a benchmark that consists of video datasets for precise $F^3$ event detection. Datasets in $F^3Set$ are characterized by their extensive scale and comprehensive detail, usually encompassing over 1,000 event types with precise timestamps and supporting multi-level granularity. Currently, $F^3Set$ contains several sports datasets, and this framework may be extended to other applications as well. We evaluated popular temporal action understanding methods on $F^3Set$, revealing substantial challenges for existing techniques. Additionally, we propose a new method, $F^3ED$, for $F^3$ event detections, achieving superior performance. The dataset, model, and benchmark code are available at https://github.com/F3Set/F3Set.

Recent studies have raised significant concerns regarding the vulnerability of Large Vision Language Models (LVLMs) to maliciously injected or perturbed input images, which can mislead their responses. Existing defense methods show that such vision attacks are sensitive to image modifications especially cropping, using majority voting across responses of modified images as corrected responses. However, these modifications often result in partial images and distort the semantics, which reduces response quality on clean images after voting. Instead of directly using responses from partial images for voting, we investigate using them to supervise (guide) the LVLM’s responses to the original images at inference time. We propose a black-box, training-free method called DPS (Defense through Partial-Perception Supervision). In this approach, the model is prompted using the responses generated by a model that perceives only a partial image. With DPS, the model can adjust its response based on partial image understanding when under attack, while confidently maintaining its original response for clean input. Empirical experiments show our method outperforms the baseline, cutting the average attack success rate by 76.3% across six datasets on three popular models.

In industrial and open-source software engineering tasks, developers often perform project-wise code editing tasks, including feature enhancement, refactoring, and bug fixing, where the leading AI models are expected to support the productivity. Hence, researchers and practitioners have proposed and adopted many LLM-based solutions to facilitate their real-world development. However, they largely suffer from the balance among predicting scope, accuracy, and efficiency. For example, solutions like Cursor achieve high accuracy only in a local editing scope while its performance drops on cross-file edits. In contrast, solutions like CoEdPilot exhibit efficiency limitations when used to predict project-wise edits.In this work, we propose TRACE (Tool-integrated RecommendAtion for Code Editing), a novel subsequent code editing solution to push the boundary of scope, accuracy, and efficiency. Our rationale lies in that code edits are triggered for either semantic or syntactic reasons. Therefore, TRACE predicts subsequent edits by interleaving neural-based induction for semantic edit prediction and tool-based deduction for syntactic edit prediction. The tools can be any IDE facilities, such as refactoring tools (e.g., rename) or linting tools (e.g., use-def), providing decent performance of deducing edit-location and edit-generation. Technically, we address the challenge of (1) when to interleave between neural-based and tool-based prediction and (2) how to further improve the performance of neural-based prediction. As for the former, we learn a neural model to detect when to invoke IDE editing tools. As for the latter, we propose a novel and fine-grained editing representation to further boost the performance of neural editing models.Our extensive experiments show that, in comparison to the state-of-the-arts such as CoEdPilot, GrACE, and CCT5, TRACE significantly improves the performance of edit location (by 43.76%) and edit generation (by 11.16%). Our simulation experiment on an interactive editing setting shows that TRACE achieves an acceptance rate 6.15% higher than Cursor. Moreover, our user study consists of 24 participants on Cursor, CoEdPilot, and TRACE, on three code editing tasks. The results show that the experimental group with TRACE achieves leading performance on cross-file global edits. In addition, we observe concerning user behaviours on how participants deal with false predictions by the tools, shedding light on the design of future code-editing tools.

Dynamic-shape tensor computation poses challenges for shape-specific compilation due to variable input dimensions. Existing compilers rely on shape samples, incurring high tuning costs and performance degradation on unseen inputs. We present Helix, a dynamic tensor compilation framework with sample-free compilation and architecture-guided optimization to achieve both compilation efficiency and shape-general performance. To avoid shape sampling, Helix constructs shape-agnostic compilation by decomposing computations across architectural layers. A bidirectional strategy combines top-down abstraction to align tensor computations with architectural hierarchies, and bottom-up kernel construction to build efficient execution strategies from reusable, architecture-aligned micro-kernels. A hybrid analyzer ensures accuracy through profiling at lower architectural levels, and achieves scalability through architecture-informed modeling at higher levels and runtime. This hierarchical design eliminates shape-specific tuning and enables shape-adaptive execution. Evaluations conducted on x86 CPUs, ARM CPUs, and NVIDIA GPUs demonstrate that Helix reduces compilation time by 174 × over the existing compilers and delivers 2.26 × and 3.29 × execution speedups over vendor libraries and dynamic-shape compilers, respectively.

Recently, the rise of code-centric Large Language Models (LLMs) has reshaped the software engineering world with low-barrier tools like Copilot that can easily generate code. However, there is no correctness guarantee for the code generated by LLMs, which suffer from the hallucination problem, and their output is fraught with risks. Besides, the end-to-end process from specification to code through LLMs is a non-transparent and uncontrolled black box. This opacity makes it difficult for users to understand and trust the generated code. Addressing these challenges is both necessary and critical. In contrast, program refinement transforms high-level specification statements into executable code while preserving correctness. Traditional tools for program refinement are primarily designed for formal methods experts and lack automation and extensibility. We apply program refinement to guide LLM and validate the LLM-generated code while transforming refinement into a more accessible and flexible framework. To initiate this vision, we propose Refine4LLM, an approach that aims to:(1) Formally refine the specifications, (2) Automatically prompt and guide the LLM using refinement calculus, (3) Interact with the LLM to generate the code, (4) Verify that the generated code satisfies the constraints, thus guaranteeing its correctness, (5) Learn and build more advanced refinement laws to extend the refinement calculus. We evaluated Refine4LLM against the state-of-the-art baselines on program refinement and LLMs benchmarks. The experiment results show that Refine4LLM can efficiently generate more robust code and reduce the time for refinement and verification.

Recent years have seen the development of LLM-based code generation. Compared to generating code in a software project, incremental code edits are empirically observed to be more frequent. The emerging code editing approaches usually formulate the problem as generating an edit based on known relevant prior edits and context. However, practical code edits can be more complicated. First, an editing session can include multiple (ir)relevant edits to the code under edit. Second, the inference of the subsequent edits is non-trivial as the scope of its ripple effect can be the whole project. In this work, we propose CoEdPilot, an LLM-driven solution to recommend code edits by discriminating the relevant edits, exploring their interactive natures, and estimating its ripple effect in the project. Specifically, CoEdPilot orchestrates multiple neural transformers to identify what and how to edit in the project regarding both edit location and edit content. When a user accomplishes an edit with an optional editing description, an Subsequent Edit Analysis first reports the most relevant files in the project with what types of edits (e.g., keep, insert, and replace) can happen for each line of their code. Next, an Edit-content Generator generates concrete edit options for the lines of code, regarding its relevant prior changes reported by an Edit-dependency Analyzer. Last, both the Subsequent Edit Analysis and the Edit-content Generator capture relevant prior edits as feedback to readjust their recommendations. We train our models by collecting over 180K commits from 471 open-source projects in 5 programming languages. Our extensive experiments show that (1) CoEdPilot can well predict the edits (i.e., predicting edit location with accuracy of 70.8%-85.3%, and the edit content with exact match rate of 41.8% and BLEU4 score of 60.7); (2) CoEdPilot can well boost existing edit generators such as GRACE and CCT5 on exact match rate by 8.57% points and BLEU4 score by 18.08. Last, our user study on 18 participants with 3 editing tasks (1) shows that CoEdPilot can be effective in assisting users to edit code in comparison with Copilot, and (2) sheds light on the future improvement of the tool design. The video demonstration of our tool is available at https://sites.google.com/view/coedpilot/home.

Data-driven techniques are promising for automatically locating and fixing bugs, which can reduce enormous time and effort for developers. However, the effectiveness of these techniques heavily relies on the quality and scale of bug datasets. Despite that emerging approaches to automatic bug dataset construction partially provide a solution for scalability, data quality remains a concern. Specifically, it remains a barrier for humans to isolate the minimal set of bug-inducing or bug-fixing changes, known as critical changes. Although delta debugging (DD) techniques are capable of extracting critical changes on benchmark datasets in academia, the efficiency and accuracy are still limited when dealing with real-world bugs, where code change dependencies could be overly complicated. In this paper, we propose C2D2, a novel delta debugging approach for critical change extraction, which estimates the probabilities of dependencies between code change elements. C2D2 considers the probabilities of dependencies and introduces a matrix-based search mechanism to resolve compilation errors (CE) caused by missing dependencies. It also provides hybrid mechanisms for flexibly selecting code change elements during the DD process. Experiments on Defect4J and a real-world regression bug dataset reveal that C2D2 is significantly more efficient than the traditional DD algorithm ddmin with competitive effectiveness, and significantly more effective and more efficient than the state-of-the-art DD algorithm ProbDD. Furthermore, compared to human-isolated critical changes, C2D2 produces the same or better critical change results in 56% cases in Defects4J and 86% cases in the regression dataset, demonstrating its usefulness in automatically extracting critical changes and saving human efforts in constructing large-scale bug datasets with real-world bugs.

Collaborative software development significantly enhances development productivity by enabling multiple contributors to work concurrently on different branches. Despite these advantages, such collaboration often increases the likelihood of causing conflicts. Resolving these conflicts brings huge challenges, primarily due to the necessity of comprehending the differences between conflicting versions. Researchers have explored various automatic conflict resolution techniques, including unstructured, structured, and learning-based approaches. However, these techniques are mostly heuristic-based or black-box in nature, which means they do not attempt to solve the root cause of the conflicts, i.e., the existence of different program behaviors exhibited by the conflicting versions.

Web applications are crucial infrastructures in the modern society, which have high demand of reliability and security. However, their frontend can be manipulable by the clients (e.g., the frontend code can be modified to bypass some validation steps), which incurs the runtime anomaly when operating the web service. Existing state-of-the-art anomaly detectors largely learn a deep learning model from the collected logs to predict abnormal logs with a probability. While effective in general, those approaches can suffer from (1) inaccuracy caused by subtle difference between the normal and abnormal/attack logs and (2) additional efforts for root cause analysis.

Relation prediction in knowledge graphs (KGs) aims at predicting missing relations in incomplete triples, whereas the dominant paradigm by KG embeddings has a limitation to predict the relation between unseen entities. This situation is called an inductive setting, which is more common in the real-world scenario. To handle this issue, implicit symbolic rules have shown great potential in capturing the inductive capability. However, it is still challenging to obtain precise representations of logic rules from KGs. The argument variability and predicate non-commutativity in symbolic rule integration make the modeling of component symbols difficult. To this end, we propose a novel inductive relation prediction model named SymRITa with a logic transformer integrating rules. SymRITa firstly extracts the subgraph, whose embeddings are captured by a graph network. Meanwhile, symbolic rule graphs in the subgraph can be generated. Then, the symbolic rules are modeled by a proposed logic transformer. Specifically, the input format based on the subgraph-based embeddings is to focus on the argument variability in symbolic rules. In addition, a conjunction attention mechanism in the logic transformer can resolve predicate non-commutativity in the symbolic rule integration process. Finally, the subgraph-based and symbol-based embeddings obtained from the previous steps are combined for the training regime, and prediction results as well as rules explaining the reasoning process are explicitly output. Extensive experiments on twelve inductive datasets show that SymRITa achieves outstanding effectiveness compared to state-of-the-art inductive baselines. Moreover, the logic rules with corresponding confidences provide an interpretable paradigm.

This is a special issue of Formal Aspects of Computing for the 28th IEEE Pacific Rim International Symposium on Dependable Computing (PRDC 2023).

Nowadays there are a wealth of devices and cameras at sports venues and facilities that collect different forms of data. Mining useful insights from such data are crucial for improving the performance of professional athletes. In this paper, we introduce a new interactive tennis analytics framework that can realistically simulate tennis matches using parameters mined from past match data and help reveal in-depth knowledge about tennis strategies. Our approach uses probabilistic model checking to formally evaluate the effectiveness of various strategies and tactics and recommend the best ones for improving players’ chances of winning. Our framework is easily understandable and actionable by players and coaches at any level. We have performed evaluations on tennis matches over the past decade to show the effectiveness of our strategy analytics framework.

Sports analytics encompasses the use of data science, AI, psychology, and IoT devices to improve sports performance, strategy, and decision-making. It involves collecting, processing, and interpreting data from various sources such as video recordings and scouting reports. The data is used to evaluate player and team performance, prevent injuries, and help coaches make informed decisions in game and training. We adopt Probabilistic Model Checking (PMC), a method commonly used in reliability analysis for complex safety systems, and explain how this method can be applied to sports strategy analytics to increase the chance of winning by taking into account the reliability of a player’s specific sub-skill sets. This paper describes how we have integrated PMC, machine learning, and computer vision to develop a new and complex system for sports strategy analytics. Finally, we discuss the vision of a new series of international sports analytics conferences (https://formal-analysis.com/isace/2023/).

Neural networks are an emerging data-driven programming paradigm widely used in many areas. Unlike traditional software systems consisting of decomposable modules, a neural network is usually delivered as a monolithic package, raising challenges for some maintenance tasks such as model restructure and re-adaption. In this work, we propose DeepArc, a novel modularization method for neural networks, to reduce the cost of model maintenance tasks. Specifically, DeepArc decomposes a neural network into several consecutive modules, each of which encapsulates consecutive layers with similar semantics. The network modularization facilitates practical tasks such as refactoring the model to preserve existing features (e.g., model compression) and enhancing the model with new features (e.g., fitting new samples). The modularization and encapsulation allow us to restructure or retrain the model by only pruning and tuning a few localized neurons and layers. Our experiments show that (1) DeepArc can boost the runtime efficiency of the state-of-the-art model compression techniques by 14.8%; (2) compared to the traditional model retraining, DeepArc only needs to train less than 20% of the neurons on average to fit adversarial samples and repair under-performing models, leading to 32.85% faster training performance while achieving similar model prediction performance.

Bugs and their fixes in the code evolution histories are important assets for many software engineering tasks such as deriving new state-of-the-art automatic bug fixing techniques. Existing bug datasets are either manually built which is difficult to grow efficiently to a scale large enough for massive data analysis, or lack of precise information of how bugs are introduced and fixed which is critical for in-depth analysis such as buggy/fixing code identification. Moreover, the types of the bugs are typically missing in the existing bug datasets, limiting the possibility of developing high-precision type-specific approaches for enterprise-level purposes. In this work, we propose BugMiner, an approach to automatically collecting bugs from code repositories by isolating the critical changes of the bugs. We also propose a learning-based approach for automating bug type classification with relatively small manual labels of bug types. We evaluate our approach regarding the precision of bug information and the efficiency of the bug-mining process with 2,082 bugs automatically mined from 100 open-source projects. We demonstrate the improved effectiveness and efficiency in bug-fixing location identification, compared to the SOTA BugBuilder, and high recall and precision in bug-inducing location identification. We also compare our learning-based bug classification approach to traditional baseline method, indicating about 17 % improvement in classification effectiveness under macro-F1.

Graph contrastive learning (GCL), aiming for an embedding space where semantically similar nodes are closer, has been widely applied in graph-structured data. Researchers have proposed many approaches to define positive and negative pairs (i.e., semantically similar and dissimilar pairs) on the graph, serving as labels to learn their embedding distances. Despite the effectiveness, those approaches usually suffer from two typical learning challenges. First, the number of candidate negative pairs is enormous. Thus, it is non-trivial to select representative ones to train the model in a more effective way. Second, the heuristics (e.g., graph views or meta-path patterns) to define positive and negative pairs are sometimes less reliable, causing considerable noise for both "labelled'' positive and negative pairs. In this work, we propose a novel sampling approach B2-Sampling to address the above challenges in a unified way. On the one hand, we use balanced sampling to select the most representative negative pairs regarding both the topological and embedding diversities. On the other hand, we use biased sampling to learn and correct the labels of the most error-prone negative pairs during the training. The balanced and biased samplings can be applied iteratively for discriminating and correcting training pairs, boosting the performance of GCL models. B2-Sampling is designed as a framework to support many known GCL models. Our extensive experiments on node classification, node clustering, and graph classification tasks show that B2-Sampling significantly improves the performance of GCL models with acceptable runtime overhead. Our website[11] https://sites.google.com/view/b2-sampling/home provides access to our codes and additional experiment results.

Researchers recently investigated to explain Graph Neural Networks (GNNs) on the access to a task-specific GNN, which may hinder their wide applications in practice. Specifically, task-specific explanation methods are incapable of explaining pretrained GNNs whose downstream tasks are usually inaccessible, not to mention giving explanations for the transferable knowledge in pretrained GNNs. Additionally, task-specific methods only consider target models' output in the label space, which are coarse-grained and insufficient to reflect the model's internal logic. To address these limitations, we consider a two-stage explanation strategy, i.e., explainers are first pretrained in a task-agnostic fashion in the representation space and then further fine-tuned in the task-specific label space and representation space jointly if downstream tasks are accessible. The two-stage explanation strategy endows post-hoc graph explanations with the applicability to pretrained GNNs where downstream tasks are inaccessible and the capacity to explain the transferable knowledge in the pretrained GNNs. Moreover, as the two-stage explanation strategy explains the GNNs in the representation space, the fine-grained information in the representation space also empowers the explanations. Furthermore, to achieve a trade-off between the fidelity and intelligibility of explanations, we propose an explanation framework based on the Information Bottleneck principle, named Explainable Graph Information Bottleneck (EGIB). EGIB subsumes the task-specific explanation and task-agnostic explanation into a unified framework. To optimize EGIB objective, we derive a tractable bound and adopt a simple yet effective explanation generation architecture. Based on the unified framework, we further theoretically prove that task-agnostic explanation is a relaxed sufficient condition of task-specific explanation, which indicates the transferability of task-agnostic explanations. Extensive experimental results demonstrate the effectiveness of our proposed explanation method.

Deep learning models are emerging to summarize source code to comment, facilitating tasks of code documentation and program comprehension. Scaled-up large language models trained on large open corpus have achieved good performance in such tasks. However, in practice, the subject code in one certain project can be specific, which may not align with the overall training corpus. Some code samples from other projects may be contradictory and introduce inconsistencies when the models try to fit all the samples. In this work, we introduce a novel approach, Adacom, to improve the performance of comment generators by on-the-fly model adaptation. This research is motivated by the observation that deep comment generators often need to strike a balance as they need to fit all the training samples. Specifically, for one certain target code $c$, some training samples $S_p$ could have made more contributions while other samples $S_o$ could have counter effects. However, the traditional fine-tuned models need to fit both $S_p$ and $S_o$ from a global perspective, leading to compromised performance for one certain target code $c$. In this context, we design Adacom to (1) detect whether the model might have a compromised performance on a target code $c$ and (2) retrieve a few helpful training samples $S_p$ that have contradictory samples in the training dataset and, (3) adapt the model on the fly by re-training the $S_p$ to strengthen the helpful samples and unlearn the harmful samples. Our extensive experiments on 7 comment generators and 4 public datasets show that (1) can significantly boost the performance of comment generation (BLEU4 score by on average 14.9\%, METEOR by 12.2\%, and ROUGE-L by 7.4\%), (2) the adaptation on one code sample is cost-effective and acceptable as an on-the-fly solution, and (3) can adapt well on out-of-distribution code samples.

Search-based software testing (SBST) has proved its effectiveness in generating test cases to achieve its defined test goals, such as branch and data-dependency coverage. However, to detect more program faults in an effective way, pre-defined goals can hardly be adaptive in diversified projects. In this work, we propose KAT, a novel knowledge-aware SBST approach to generate on-demand assertions in the program under test (PUT) based on its used APIs. KAT constructs an API knowledge graph from the API documentation to derive the constraints that the client codes need to satisfy. Each constraint is instrumented into the PUT as a program branch, serving as a test goal to guide SBST to detect faults. We evaluate KAT with two baselines (i.e., EvoSuite and Catcher) with a close-world and an open-world experiment to detect API bugs. The close-world experiment shows that KAT outperforms the baselines in the F1-score (0.55 vs. 0.24 and 0.30) to detect API-related bugs. The open-world experiment shows that KAT can detect 59.64% and 9.05% more bugs than the baselines in practice.

A deep classifier is usually trained to (i) learn the numeric representation vector of samples and (ii) classify sample representations with learned classification boundaries. Time-travelling visualization, as an explainable AI technique, is designed to transform the model training dynamics into an animation of canvas with colorful dots and territories. Despite that the training dynamics of the high-level concepts such as sample representations and classification boundaries are now observable, the model developers can still be overwhelmed by tens of thousands of moving dots across hundreds of training epochs (i.e., frames in the animation), which makes them miss important training events.

Over the past few years, SMT string solvers have found their applications in an increasing number of domains, such as program analyses in mobile and Web applications, which require the ability to reason about string values. A series of research has been carried out to find quality issues of string solvers in terms of its correctness and performance. Yet, none of them has considered the performance regressions happening across multiple versions of a string solver. To fill this gap, in this paper, we focus on solver performance regressions (SPRs), i.e., unintended slowdowns introduced during the evolution of string solvers. To this end, we developSPRFinderto not only generate test cases demonstrating SPRs, but also localize the probable causes of them, in terms of commits. We evaluated the effectiveness ofSPRFinderon three state-of-the-art string solvers, i.e., Z3Seq, Z3Str3, and CVC4. The results demonstrate thatSPRFinderis effective in generating SPR-inducing test cases and also able to accurately locate the responsible commits. Specifically, the average running time on the target versions is 13.2× slower than that of the reference versions. Besides, we also conducted the first empirical study to peek into the characteristics of SPRs, including the impact of random seed configuration for SPR detection, understanding the root causes of SPRs, and characterizing the regression test cases through case studies. Finally, we highlight that 149 unique SPR-inducing commits were discovered in total bySPRFinder, and 27of them have been confirmed by the corresponding developers.

Understanding how the predictions of deep learning models are formed during the training process is crucial to improve model performance and fix model defects, especially when we need to investigate nontrivial training strategies such as active learning, and track the root cause of unexpected training results such as performance degeneration. In this work, we propose a time-travelling visual solution DeepVisualInsight (DVI), aiming to manifest the spatio-temporal causality while training a deep learning image classifier. The spatio-temporal causality demonstrates how the gradient-descent algorithm and various training data sampling techniques can influence and reshape the layout of learnt input representation and the classification boundaries in consecutive epochs. Such causality allows us to observe and analyze the whole learning process in the visible low dimensional space. Technically, we propose four spatial and temporal properties and design our visualization solution to satisfy them. These properties preserve the most important information when projecting and inverse-projecting input samples between the visible low-dimensional and the invisible high-dimensional space, for causal analyses. Our extensive experiments show that, comparing to baseline approaches, we achieve the best visualization performance regarding the spatial/temporal properties and visualization efficiency. Moreover, our case study shows that our visual solution can well reflect the characteristics of various training scenarios, showing good potential of DVI as a debugging tool for analyzing deep learning training processes.

Time-travelling visualization answers how the predictions of a deep classifier are formed during the training. It visualizes in two or three dimensional space how the classification boundaries and sample embeddings are evolved during training. In this work, we propose TimeVis, a novel time-travelling visualization solution for deep classifiers. Comparing to the state-of-the-art solution DeepVisualInsight (DVI), TimeVis can significantly (1) reduce visualization errors for rendering samples’ travel across different training epochs, and (2) improve the visualization efficiency. To this end, we design a technique called temporality spatialization, which unifies the spatial relation (e.g., neighbouring samples in single epoch) and temporal relation (e.g., one identical sample in neighbouring training epochs) into one high-dimensional topological complex. Such spatio-temporal complex can be used to efficiently train one visualization model to accurately project and inverse-project any high and low dimensional data across epochs. Our extensive experiment shows that, in comparison to DVI, TimeVis not only is more accurate to preserve the visualized time-travelling semantics, but 15X faster in visualization efficiency, achieving a new state-of-the-art in time-travelling visualization.

Bug datasets lay significant empirical and experimental foundation for various SE/PL researches such as fault localization, software testing, and program repair. Current well-known datasets are constructed manually, which inevitably limits their scalability, representativeness, and the support for the emerging data-driven research.

Trained with a sufficiently large training and testing dataset, Deep Neural Networks (DNNs) are expected to generalize. However, inputs may deviate from the training dataset distribution in real deployments. This is a fundamental issue with using a finite dataset, which may lead deployed DNNs to mis-predict in production.

Deep metric learning (DML) learns a generalizable embedding space where the representations of semantically similar samples are closer. Despite achieving good performance, the state-of-the-art models still suffer from the generalization errors such as farther similar samples and closer dissimilar samples in the space. In this work, we design an empirical influence function (EIF), a debugging and explaining technique for the generalization errors of state-of-the-art metric learning models. EIF is designed to efficiently identify and quantify how a subset of training samples contributes to the generalization errors. Moreover, given a user-specific error, EIF can be used to relabel a potentially noisy training sample as mitigation. In our quantitative experiment, EIF outperforms the traditional baseline in identifying more relevant training samples with statistical significance and 33.5% less time. In the field study on well-known datasets such as CUB200, CARS196, and InShop, EIF identifies 4.4%, 6.6%, and 17.7% labelling mistakes, indicating the direction of the DML community to further improve the model performance. Our code is available at https://github.com/lindsey98/Influencefunctionmetric_learning.

In this work, we introduce a tool, RegMiner, to automate the process of collecting replicable regression bugs from a set of Git repositories. In the code commit history, RegMiner searches for regressions where a test can pass a regression-fixing commit, fail a regressioninducing commit, and pass a previous working commit again. Technically, RegMiner (1) identifies potential regression-fixing commits from the code evolution history, (2) migrates the test and its code dependencies in the commit over the history, and (3) minimizes the compilation overhead during the regression search. Our experients show that RegMiner can successfully collect 1035 regressions over 147 projects in 8 weeks, creating the largest replicable regression dataset within the shortest period, to the best of our knowledge. In addition, our experiments further show that (1) RegMiner can construct the regression dataset with very high precision and acceptable recall, and (2) the constructed regression dataset is of high authenticity and diversity. The source code of RegMiner is available at https://github.com/SongXueZhi/RegMiner, the mined regression dataset is available at https://regminer.github.io/, and the demonstration video is available at https://youtu.be/yzcM9Y4unok.

The widespread adoption of Deep Neural Networks (DNNs) in important domains raises questions about the trustworthiness of DNN outputs. Even a highly accurate DNN will make mistakes some of the time, and in settings like self-driving vehicles these mistakes must be quickly detected and properly dealt with in deployment. Just as our community has developed effective techniques and mechanisms to monitor and check programmed components, we believe it is now necessary to do the same for DNNs. In this paper we present DNN self-checking as a process by which internal DNN layer features are used to check DNN predictions. We detail SelfChecker, a self-checking system that monitors DNN outputs and triggers an alarm if the internal layer features of the model are inconsistent with the final prediction. SelfChecker also provides advice in the form of an alternative prediction. We evaluated SelfChecker on four popular image datasets and three DNN models and found that SelfChecker triggers correct alarms on 60.56% of wrong DNN predictions, and false alarms on 2.04% of correct DNN predictions. This is a substantial improvement over prior work (SelfOracle, Dissector, and ConfidNet). In experiments with self-driving car scenarios, SelfChecker triggers more correct alarms than SelfOracle for two DNN models (DAVE-2 and Chauffeur) with comparable false alarms. Our implementation is available as open source.

Fuzzing has been a widely-used technique for discovering software vulnerabilities. Many existing fuzzers leverage coverage-feedback to evolve seeds to maximize (optimize) program branch coverage. Recently, some techniques propose to train deep learning models to predict the branch coverage of an arbitrary input. Those techniques have proved their success in improving coverage and discovering bugs under different experimental settings. However, deep learning models, usually as a black magic box, are notoriously lack of explanation. Moreover, their performance can be sensitive to the collected runtime coverage information for training, indicating potentially unstable performance. To this end, in this work we conduct a systematic and extensive empirical study on 4 types of deep learning models across 6 projects to reproduce the actual performance of deep learning fuzzers, analyze the advantages and disadvantages of deep learning in the process of fuzzing applications, and explore the future direction of the combination of the two. Our empirical results reveal that the deep learning models can only be effective in very limited scenarios, which is largely restrained by training data imbalance, dependant labels, model over-generalization, and the insufficient expressiveness of the state-of-the-art models. Consequently, the estimated gradients by the models to cover a branch can be less helpful in many scenarios.

Search-based software testing (SBST) generates tests using search algorithms guided by measurements gauging how far a test case is away from exercising a coverage goal. The effectiveness of SBST largely depends on the continuity and monotonicity of the fitness landscape decided by these measurements and the search operators. Unfortunately, the fitness landscape is challenging when the function under test takes object inputs, as classical measurement hardly provide guidance for constructing legitimate object inputs. To overcome this problem, we propose test seeds, i.e., test code skeletons of legitimate objects which enable the use of classical measurements. Given a target branch in a function under test, we first statically analyze the function to build an object construction graph that captures the relation between the operands of the target method and the states of their relevant object inputs. Based on the graph, we synthesize test template code where each "slot" is a mutation point for the search algorithm. This approach can be seamlessly integrated with existing SBST algorithms, and we implemented EvoObj on top of EvoSuite. Our experiments show that EvoObj outperforms EvoSuite with statistical significance on 2750 methods over 103 open source Java projects using state-of-the-art SBST algorithms.

Adversarial Examples threaten to fool deep learning models to output erroneous predictions with high confidence. Optimization-based methods for constructing such samples have been extensively studied. While being effective in terms of aggression, they typically lack clear interpretation and constraint about their underlying generation process, which thus hinders us from leveraging the produced adversarial samples for model protection in the reverse direction. Hence, we expect them to repair bugs in the pre-trained models by produced additional training data equipped with strong attack ability rather than time-consuming full re-training from scratch. To address these issues, we first study the black-box behaviors and the intrinsic deficiency of neighborhood information in previous optimization-based adversarial attacks and defenses, respectively. Then we introduce a new method dubbed FeaCP, which uses correct predicted samples in disjoint classes to guide the generation of more explainable adversarial samples in the ambiguous region around the decision boundary instead of uncontrolled “blind spots”, via convex combination in a feature component-wise manner which takes the individual importance of feature ingredients into account. Our method incorporates the prior fact that for well-separated samples, the path connecting them would go through model's decision-boundary that lies in a low-density region, however, wherein adversarial examples are spread with high probability, thus having an impact on the ultimate trained model. In our work, the path is constructed by proposed inhomogeneous feature-wise convex interpolation rather than operating on sample-wise level, limiting the search space of FeaCP to obtain an adaptive neighborhood. Finally, we provide detailed insights and extend our method to adversarial fine-tuning using vicinity distribution to optimize the approximated decision boundary, and validate the significance of our FeaCP to model performance. The experimental results show that our method provides competitive performance on various datasets and networks.

Regression faults, which make working code stop functioning, are often introduced when developers make changes to the software. Many regression fault localization techniques have been proposed. However, issues like inaccuracy and lack of explanation are still obstacles for their practical application. In this work, we propose a trace-based approach to identifying not only where the root cause of a regression bug lies, but also how the defect is propagated to its manifestation as the explanation. In our approach, we keep the trace of original correct version as reference and infer the faulty steps on the trace of regression version so that we can build a causality graph of how the defect is propagated. To this end, we overcomes two technical challenges. First, we align two traces derived from two program versions by extending state-of-the-art trace alignment technique for regression fault with novel relaxation technique. Second, we construct causality graph (i.e., explanation) by adopting a technique calledalignment slicing and mendingto isolate the failure-inducing changes and explain the failure. Our comparative experiment with the state-of-the-art techniques including dynamic slicing, delta-debugging, and symbolic execution on 24 real-world regressions shows that (1) our approach is more accurate on isolating the failure-inducing changes, (2) the generated explanation requires acceptable manual effort to inspect, and (3) our approach requires lower runtime overhead. In addition, we also conduct an applicability experiment based on Defects4J bug repository, showing the potential limitations of our trace-based approach and providing guidance for its practical use.

Smart contracts are Turing-complete programs that execute on the infrastructure of the blockchain, which often manage valuable digital assets. Solidity is one of the most popular programming languages for writing smart contracts on the Ethereum platform. Like traditional programs, smart contracts may contain vulnerabilities. Unlike traditional programs, smart contracts cannot be easily patched once they are deployed. It is thus important that smart contracts are tested thoroughly before deployment. In this work, we present an adaptive fuzzer for smart contracts on the Ethereum platform called sFuzz. Compared to existing Solidity fuzzers, sFuzz combines the strategy in the AFL fuzzer and an efficient lightweight multi-objective adaptive strategy targeting those hard-to-cover branches. sFuzz has been applied to more than 4 thousand smart contracts and the experimental results show that (1) sFuzz is efficient, e.g., two orders of magnitude faster than state-of-the-art tools; (2) sFuzz is effective in achieving high code coverage and discovering vulnerabilities; and (3) the different fuzzing strategies in sFuzz complement each other.

In Search-based Software Testing (SBST), test generation is guided by fitness functions that estimate how close a test case is to reach an uncovered test goal (e.g., branch). A popular fitness function estimates how close conditional statements are to evaluating to true or false, i.e., the branch distance. However, when conditions read Boolean variables (e.g., if(x && y)), the branch distance provides no gradient for the search, since a Boolean can either be true or false. This flag problem can be addressed by transforming individual procedures such that Boolean flags are replaced with numeric comparisons that provide better guidance for the search. Unfortunately, defining a semantics-preserving transformation that is applicable in an interprocedural case, where Boolean flags are passed around as parameters and return values, is a daunting task. Thus, it is not yet supported by modern test generators.

Reentrancy bugs, one of the most severe vulnerabilities in smart contracts, have caused huge financial loss in recent years. Researchers have proposed many approaches to detecting them. However, empirical studies have shown that these approaches suffer from undesirable false positives and false negatives, when the code under detection involves the interaction between multiple smart contracts.

Locating vulnerabilities is an important task for security auditing, exploit writing, and code hardening. However, it is challenging to locate vulnerabilities in binary code, because most program semantics (e.g., boundaries of an array) is missing after compilation. Without program semantics, it is difficult to determine whether a memory access exceeds its valid boundaries in binary code. In this work, we propose an approach to locate vulnerabilities based on memory layout recovery. First, we collect a set of passed executions and one failed execution. Then, for passed and failed executions, we restore their program semantics by recovering fine-grained memory layouts based on the memory addressing model. With the memory layouts recovered in passed executions as reference, we can locate vulnerabilities in failed execution by memory layout identification and comparison. Our experiments show that the proposed approach is effective to locate vulnerabilities on 24 out of 25 DARPA’s CGC programs (96%), and can effectively classifies 453 program crashes (in 5 Linux programs) into 19 groups based on their root causes.

Concolic testing integrates concrete execution (e.g., random testing) and symbolic execution for test case generation. It is shown to be more cost-effective than random testing or symbolic execution sometimes. A concolic testing strategy is a function which decides when to apply random testing or symbolic execution, and if it is the latter case, which program path to symbolically execute. Many heuristics-based strategies have been proposed. It is still an open problem what is the optimal concolic testing strategy. In this work, we make two contributions towards solving this problem. First, we show the optimal strategy can be defined based on the probability of program paths and the cost of constraint solving. The problem of identifying the optimal strategy is then reduced to a model checking problem of Markov Decision Processes with Costs. Secondly, in view of the complexity in identifying the optimal strategy, we design a greedy algorithm for approximating the optimal strategy. We conduct two sets of experiments. One is based on randomly generated models and the other is based on a set of C programs. The results show that existing heuristics have much room to improve and our greedy algorithm often outperforms existing heuristics.

Dynamic slicing is a common way of identifying the root cause when a program fault is revealed. With the dynamic slicing technique, the programmers can follow data and control flow along the program execution trace to the root cause. However, the technique usually fails to work on omission bugs, i.e., the faults which are caused by missing executing some code. In many cases, dynamic slicing over-skips the root cause when an omission bug happens, leading the debugging process to a dead end. In this work, we conduct an empirical study on the omission bugs in the Defects4J bug repository. Our study shows that (1) omission bugs are prevalent (46.4%) among all the studied bugs; (2) there are repeating patterns on causes and fixes of the omission bugs; (3) the patterns of fixing omission bugs serve as a strong hint to break the slicing dead end. Based on our findings, we train a neural network model on the omission bugs in Defects4J repository to recommend where to approach when slicing can no long work. We conduct an experiment by applying our approach on 3193 mutated omission bugs which slicing fails to locate. The results show that our approach outperforms random benchmark on breaking the dead end and localizing the mutated omission bugs (63.8% over 2.8%).

Software debugging has long been regarded as a time and effort consuming task. In the process of debugging, developers usually need to manually inspect many program steps to see whether they deviate from their intended behaviors. Given that intended behaviors usually exist nowhere but in human mind, the automation of debugging turns out to be extremely hard, if not impossible. In this work, we propose a feedback-based debugging approach, which (1) builds on light-weight human feedbacks on a buggy program and (2) regards the feedbacks as partial program specification to infer suspicious steps of the buggy execution. Given a buggy program, we record its execution trace and allow developers to provide light-weight feedback on trace steps. Based on the feedbacks, we recommend suspicious steps on the trace. Moreover, our approach can further learn and approximate bug-free paths, which helps reduce required feedbacks to expedite the debugging process. We conduct an experiment to evaluate our approach with simulated feedbacks on 3409 mutated bugs across 3 open source projects. The results show that our feedback-based approach can detect 92.8% of the bugs and 65% of the detected bugs require less than 20 feedbacks. In addition, we implement our proof-of-concept tool, Microbat, and conduct a user study involving 16 participants on 3 debugging tasks. The results show that, compared to the participants using the baseline tool, Whyline, the ones using Microbat can spend on average 55.8% less time to locate the bugs.

In this paper, we propose an approach to detecting project-specific recurring designs in code base and abstracting them into design templates as reuse opportunities. The mined templates allow programmers to make further customization for generating new code. The generated code involves the code skeleton of recurring design as well as the semi-implemented code bodies annotated with comments to remind programmers of necessary modification. We implemented our approach as an Eclipse plugin called MICoDe. We evaluated our approach with a reuse simulation experiment and a user study involving 16 participants. The results of our simulation experiment on 10 open source Java projects show that, to create a new similar feature with a design template, (1) on average 69% of the elements in the template can be reused and (2) on average 60% code of the new feature can be adopted from the template. Our user study further shows that, compared to the participants adopting the copy-paste-modify strategy, the ones using MICoDe are more effective to understand a big design picture and more efficient to accomplish the code reuse task.

Architectural refactorings can contain hundreds of steps and experienced developers could carry them out over several weeks. Moreover, developers need to explore a correct sequence of refactorings steps among many more incorrect alternatives. Thus, carrying out architectural refactorings is costly, risky, and challenging. In this paper, we present Refactoring Navigator: a tool-supported and interactive recommendation approach for aiding architectural refactoring. Our approach takes a given implementation as the starting point, a desired high-level design as the target, and iteratively recommends a series of refactoring steps. Moreover, our approach allows the user to accept, reject, or ignore a recommended refactoring step, and uses the user's feedback in further refactoring recommendations. We evaluated the effectiveness of our approach and tool using a controlled experiment and an industrial case study. The controlled experiment shows that the participants who used Refactoring Navigator accomplished their tasks in 77.4% less time and manually edited 98.3% fewer lines than the control group. The industrial case study suggests that Refactoring Navigator has the potential to help with architectural refactorings in practice.

Developers often need to modify pasted code when programming with copy-and-paste practice. Some modifications on pasted code could involve lots of editing efforts, and any missing or wrong edit could incur bugs. In this paper, we propose a clone-based and interactive approach to recommending where and how to modify the pasted code. In our approach, we regard clones of the pasted code as the results of historical copy-and-paste operations and their differences as historical modifications on the same piece of code. Our approach first retrieves clones of the pasted code from a clone repository and detects syntactically complete differences among them. Then our approach transfers each clone difference into a modification slot on the pasted code, suggests options for each slot, and further mines modifying regulations from the clone differences. Based on the mined modifying regulations, our approach dynamically updates the suggested options and their ranking in each slot according to developer's modifications on the pasted code. We implement a proof-of-concept tool CCDemon based on our approach and evaluate its effectiveness based on code clones detected from five open source projects. The results show that our approach can identify 96.9% of the to-be-modified positions in pasted code and suggest 75.0% of the required modifications. Our human study further confirms that CCDemon can help developers to accomplish their modifications of pasted code more efficiently.

Clone detectors find similar code fragments (i.e., instances of code clones) and report large numbers of them for industrial systems. To maintain or manage code clones, developers often have to investigate differences of multiple cloned code fragments. However,existing program differencing techniques compare only two code fragments at a time. Developers then have to manually combine several pairwise differencing results. In this paper, we present an approach to automatically detecting differences across multiple clone instances. We have implemented our approach as an Eclipse plugin and evaluated its accuracy with three Java software systems. Our evaluation shows that our algorithm has precision over 97.66% and recall over 95.63% in three open source Java projects. We also conducted a user study of 18 developers to evaluate the usefulness of our approach for eight clone-related refactoring tasks. Our study shows that our approach can significantly improve developers’performance in refactoring decisions, refactoring details, and task completion time on clone-related refactoring tasks. Automatically detecting differences across multiple clone instances also opens opportunities for building practical applications of code clones in software maintenance, such as auto-generation of application skeleton, intelligent simultaneous code editing.

Code clones have to be made explicit and be managed in software maintenance. Researchers have developed many clone detection tools to detect and analyze code clones in software systems. These tools report code clones as similar code fragments in source files. However, clone-related maintenance tasks (e.g., refactorings) often involve a group of code clones appearing in larger syntactic context (e.g., code clones in sibling classes or code clones calling similar methods). Given a list of low-level code-fragment clones, developers have to manually summarize from bottom up low-level code clones that are relevant to the syntactic context of a maintenance task. In this paper, we present a clone summarization technique to summarize code clones with respect to their common syntactic context. The clone summarization allows developers to locate and maintain code clones in a top-down manner by type hierarchy and usage dependencies. We have implemented our approach in the Clonepedia tool and conducted a user study on JHotDraw with 16 developers. Our results show that Clonepedia users can better locate and refactor code clones, compared with developers using the Clone Detective tool.