Information and Software Technology最新文献

{"title":"SemiRALD: A semi-supervised hybrid language model for robust Anomalous Log Detection","authors":"Yicheng Sun ,&nbsp;Jacky Wai Keung ,&nbsp;Zhen Yang ,&nbsp;Shuo Liu ,&nbsp;Hi Kuen Yu","doi":"10.1016/j.infsof.2025.107743","DOIUrl":"10.1016/j.infsof.2025.107743","url":null,"abstract":"<div><h3>Context:</h3><div>Deep learning-based Anomalous Log Detection (DALD) tools are critical for software reliability, but current approaches face challenges, including information loss during log parsing, reliance on large labeled datasets, and fragility in low-resource scenarios.</div></div><div><h3>Objective:</h3><div>To overcome the above limitations, we propose SemiRALD, a semi-supervised learning-based robust ALD approach that leverages Large Language Model (LLM) for log parsing, enhancing both flexibility and accuracy. It utilizes a hybrid language model to repeatedly fit the samples with generate pseudo-labels, thereby training DALD models with limited resources and facilitating efficient anomaly detection tasks.</div></div><div><h3>Method:</h3><div>In detail, SemiRALD utilizes ChatGPT and in-context learning for automated log parsing, thereby improving the log integrity during log parsing. Subsequently, it harnesses a semi-supervised learning framework and our proposed hybrid language model to remedy the performance degeneration caused by low-resource restriction in practice. Semi-supervised learning requires only a small amount of labeled data throughout the entire process, while the hybrid language model is built on the architecture of RoBERTa and an attention-based BiLSTM.</div></div><div><h3>Results:</h3><div>Experiments on the HDFS and BGL datasets demonstrate that SemiRALD achieves an average F1-score improvement of 7.3% and 8.2%, respectively, over seven benchmark models. On small-scale datasets (0.1% of the original size), SemiRALD outperforms competitors by 31.4% and 46.0% in F1-score, respectively. Its consistent performance across diverse datasets highlights its generalizability and robustness.</div></div><div><h3>Conclusion:</h3><div>SemiRALD is capable of handling anomaly detection tasks in both large-scale and low-resource datasets, delivering significant advancements in anomaly log detection and offering robust, adaptable solutions to address prevalent challenges in the field of software reliability engineering.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107743"},"PeriodicalIF":3.8,"publicationDate":"2025-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834794","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"RaxCS: Towards cross-language code summarization with contrastive pre-training and retrieval augmentation","authors":"Kaiyuan Yang ,&nbsp;Junfeng Wang ,&nbsp;Zihua Song","doi":"10.1016/j.infsof.2025.107741","DOIUrl":"10.1016/j.infsof.2025.107741","url":null,"abstract":"<div><h3>Context:</h3><div>Code summarization is the task of generating a concise natural language description of the code snippet. Recent efforts have been made to boost the performance of code summarization language from various perspectives, e.g., retrieving external information or introducing large transformer-based models, and thus has achieved promising performance for one specific programming language. While dealing with rapidly expanded cross-language source code datasets, existing approaches suffer from two issues, (1) the difficulty of building a universe code representation for multiple languages; (2) less-well performance for low-resource language.</div></div><div><h3>Objective:</h3><div>To cope with these issues, we propose a novel code summarization approach named RaxCS, which aims to perform code summarization across multiple languages and improve accuracy for low-resource languages by leveraging cross-language knowledge.</div></div><div><h3>Methods:</h3><div>We exploit the pre-trained models with the contrastive learning objective to build a unified code representation towards multiple languages. To fully mine the external knowledge across programming languages, we design a hybrid retrieval module to search functionally equivalent code and its corresponding comment to serve as preliminary information. Finally, we employ a decode-only transformer model to fuse contextual information, which guides the process of generating summaries.</div></div><div><h3>Results:</h3><div>Extensive experiments demonstrate (1) RaxCS outperforms the state-of-the-art on cross-language code summarization (i.e., RaxCS scores 4.39% higher in terms of BLEU metric and 8.65% in terms of BERTScore). (2) For low-resource languages, RaxCS can boost the code summarization performance by a significant magnification (e.g., 6.93% in terms of BLEU for ruby) with cross-language retrieval.</div></div><div><h3>Conclusion:</h3><div>This paper introduces a cross-language code summarization model, which utilizes contrastive pre-training and cross-language retrieval. Both are beneficial for incorporating cross-language knowledge to advance code summarization performance. The experimental results demonstrate that RaxCS is effective in generating accurate code summaries, particularly for low-resource languages.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107741"},"PeriodicalIF":3.8,"publicationDate":"2025-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143820618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Validation of information architecture: Cross-methodological comparison of tree testing variants and prototype user testing","authors":"Eduard Kuric ,&nbsp;Peter Demcak ,&nbsp;Matus Krajcovic","doi":"10.1016/j.infsof.2025.107740","DOIUrl":"10.1016/j.infsof.2025.107740","url":null,"abstract":"<div><h3>Context:</h3><div>Tree testing is an established user testing method applied by software professionals to validate that an information architecture is logically navigable by users. We identify a methodological gap caused by previously unexamined non-uniformity between tree testing methods and software.</div></div><div><h3>Objective:</h3><div>To reveal the role of the user interface representations in tree testing, this research compares the results of 3 commonly-used tree testing variants. To assess how indicative they are of the user’s interaction with an information architecture implemented in an actual user interface, and to issue methodological recommendations, comparison with varied high-fidelity prototypes was performed.</div></div><div><h3>Methods:</h3><div>Two between-subject studies were conducted to obtain a new dataset of users navigating an information architecture in tree testing and in interactive user interface prototypes. Data from 180 participants and 1800 task completions between 6 experimental conditions—3 tree testing and 3 prototype user interface variants—was evaluated quantitatively and qualitatively.</div></div><div><h3>Results:</h3><div>Significant differences were found between results yielded by different tree testing method variants, and in how well they approximate user navigation in the same information architecture in high-fidelity prototypes. Implications for selection of the tree testing variant are proposed in the context of evaluated information architecture, with plausible broader applicability for tree testing methodology. Evidence supports the tree testing variant with highest visibility of previous navigation choices and direct controls over their reversal as the most accurate.</div></div><div><h3>Conclusion:</h3><div>Presented findings can contribute to the design of software information architecture based on more accurate early validation, owing to tree testing that simulates less artificial user behavior more reflective of the user’s navigation in the eventual user interface. We hope this will further the discussion and research leading to more holistic tree testing methodologies in the future.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107740"},"PeriodicalIF":3.8,"publicationDate":"2025-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143817124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"VDMAF: Cross-language source code vulnerability detection using multi-head attention fusion","authors":"Yang Li ,&nbsp;Qin Luo ,&nbsp;Peng Wu ,&nbsp;Hongdi Zheng","doi":"10.1016/j.infsof.2025.107739","DOIUrl":"10.1016/j.infsof.2025.107739","url":null,"abstract":"<div><h3>Context:</h3><div>Detecting potential vulnerabilities is critical for ensuring the stability and reliability of software systems. Traditional static detection methods fall short in accuracy and efficiency. Furthermore, existing deep learning-based vulnerability detection models typically rely on single sequence or graph embedding methods, neglecting the semantic and structured information present in the code. With the diversification of software development environments, systems often involve multiple programming languages. This limits the effectiveness of existing vulnerability detection methods when handling cross-language code.</div></div><div><h3>Objective:</h3><div>To solve these problems, we propose a more effective and general vulnerability detection framework, VDMAF(Cross-Language Source Code Vulnerability Detection Using Multi-Head Attention Fusion).</div></div><div><h3>Methods:</h3><div>The method extracts unified and standardized feature representations. It uses a multi-head attention module to fuse sequence features and graph structural features. First, an improved global consistent labeling mechanism is introduced, which improves data representation through threshold-based label augmentation. Second, the method uses sequence embedding to extract local semantic features of the code. The code is converted into a unified, standardized graph structure. Then, a graph neural network is used to extract features. Finally, the sequence and graph features are fused using the multi-head attention module, followed by classification with a bidirectional LSTM-based recurrent neural network.</div></div><div><h3>Results:</h3><div>VDMAF has been evaluated on three vulnerability datasets across different programming languages and granularities, demonstrating better performance across all metrics compared to baseline models, with F1 scores of 98.9%, 65.3%, and 56.8%.</div></div><div><h3>Conclusion:</h3><div>The proposed VDMAF outperforms state-of-the-art models, exhibiting better generality and scalability, thus showing greater potential in vulnerability detection tasks.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107739"},"PeriodicalIF":3.8,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unraveling the pain points of domain modeling","authors":"Isadora Valle ,&nbsp;Tiago Prince Sales ,&nbsp;Eduardo Guerra ,&nbsp;Maya Daneva ,&nbsp;Renata Guizzardi ,&nbsp;Luiz Olavo Bonino da Silva Santos ,&nbsp;Henderik A. Proper ,&nbsp;Giancarlo Guizzardi","doi":"10.1016/j.infsof.2025.107736","DOIUrl":"10.1016/j.infsof.2025.107736","url":null,"abstract":"<div><div>Conceptual models offer numerous benefits but require significant investments, requiring modelers to strive to balance costs and benefits. Understanding the modeling process and the frustrations experienced by modelers can provide valuable insights for this assessment. While research acknowledges certain instances of modelers’ dissatisfaction, its scope often limits detailed examination. This study seeks to identify and analyze the main pain points associated with domain modeling through a five-phase empirical study using a multi-method approach. We identified <strong>71</strong> pain points, synthesized them to <strong>41</strong>, and prioritized <strong>16</strong> as the most significant and prevalent in domain modeling. We then refined, documented, and exemplified the prioritized pain points, analyzed their potential causes, and discussed their practical implications. Our findings provide valuable insights for improving modelers’ experiences and optimizing the modeling process.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107736"},"PeriodicalIF":3.8,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143843731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Don’t settle for the first! How many GitHub Copilot solutions should you check?","authors":"Julian Oertel ,&nbsp;Jil Klünder ,&nbsp;Regina Hebig","doi":"10.1016/j.infsof.2025.107737","DOIUrl":"10.1016/j.infsof.2025.107737","url":null,"abstract":"<div><h3>Context:</h3><div>With the integration of generative artificial intelligence (GenAI) tools such as GitHub Copilot into development processes, developers can be supported when writing code.</div></div><div><h3>Objectives:</h3><div>As GitHub Copilot has a feature to provide up to ten solutions at once, we explore, how developers should approach those solutions with the goal of providing recommendations to achieve suitable trade-offs in finding correct solutions and checking solutions.</div></div><div><h3>Methods:</h3><div>In this study, we analyze a total of 2025 coding problems provided by LeetCode and 17<!--> <!-->048 solutions to solve these problems generated by GitHub Copilot in Python. We focus on three key issues: firstly, whether it is beneficial to consider multiple solutions; secondly, the impact of the position of a solution; and thirdly, the number of solutions that should be checked by a developer.</div></div><div><h3>Results:</h3><div>Overall, our results point to the following observations: (1) solutions are not less likely to be correct if they appear at later positions; (2) when looking for a solution to a common problem, checking four to five solutions is generally enough; (3) novel or difficult problems are unlikely to be solved by GitHub Copilot; (4) skipping the first solution is advised when considering only one solution, as the first solution is less likely to be correct; and (5) checking all solutions is necessary to not miss correct solutions, but the effort is usually not justified.</div></div><div><h3>Conclusion:</h3><div>Based on our study, we conclude that there is potential for improvement in better supporting developers. For instance, there are few cases where ten generated solutions provide more value than fewer solutions. Depending on the use scenario, it could be more useful if GitHub Copilot allowed developers to request a single, comprehensive solution.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107737"},"PeriodicalIF":3.8,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834796","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Extensive mutation for testing of word sense disambiguation models","authors":"Deping Zhang ,&nbsp;Zhaohui Yang ,&nbsp;Xiang Huang ,&nbsp;Yanhui Li","doi":"10.1016/j.infsof.2025.107734","DOIUrl":"10.1016/j.infsof.2025.107734","url":null,"abstract":"<div><h3>Context:</h3><div>Word sense disambiguation (WSD) models are extensively utilized in various translation and question-answering systems. Assessing the WSD capability of these models aids in their improvement and enhances their dependability. Recently, researchers have introduced the concept of “mutation” to induce WSD errors in machine translation systems to evaluate their WSD ability.</div></div><div><h3>Objective:</h3><div>Inspired by the recent research, this study aims to extend types of mutations and check their potential application in testing WSD models to check whether these mutations can effectively provoke WSD errors.</div></div><div><h3>Method:</h3><div>We have designed and implemented nine innovative types of mutations focusing on words, phrases, and sentence structure for the sentence in WSD testing. Based on these extensive mutations, we have proposed a WSD testing framework that utilizes large language models to prompt sentence mutations and assess the disambiguation capability of WSD models.</div></div><div><h3>Results:</h3><div>In our research, we have conducted experiments using five widely recognized test sets for WSD tasks under five widely used WSD models. The experimental results show that (a) our testing framework can produce correct mutations for nine proposed mutations, and (b) the newly developed mutations have been shown to successfully trigger a substantial number of factual and unique WSD errors.</div></div><div><h3>Conclusions:</h3><div>The new types of mutations we designed can effectively be applied in mutation-based WSD testing. This suggests that by exploring more types of mutations, more WSD errors can be triggered.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107734"},"PeriodicalIF":3.8,"publicationDate":"2025-04-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143800724","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Copiloting the future: How generative AI transforms Software Engineering","authors":"Leonardo Banh ,&nbsp;Florian Holldack ,&nbsp;Gero Strobel","doi":"10.1016/j.infsof.2025.107751","DOIUrl":"10.1016/j.infsof.2025.107751","url":null,"abstract":"<div><h3><strong>Context</strong></h3><div>With rapid technological advancements, artificial intelligence (AI) has become integral to various sectors. Generative AI (GenAI) tools like ChatGPT or GitHub Copilot, with their unique content creation capabilities, pose transformative potential in Software Engineering by offering new ways to optimize software development processes. However, the integration into current processes also presents challenges that require a sociotechnical analysis to effectively realize GenAI's potential.</div></div><div><h3><strong>Objective</strong></h3><div>This study investigates how GenAI can be leveraged in the domain of Software Engineering, exploring its action potentials and challenges to help businesses and developers optimize the adoption of this technology in their workflows.</div></div><div><h3><strong>Method</strong></h3><div>We performed a qualitative study and collected data from expert interviews with eighteen professionals working in Software Engineering-related roles. Data analysis followed the principles of Grounded Theory to analyze how GenAI supports developers' goals, aligns with organizational practices, and facilitates integration into existing routines.</div></div><div><h3><strong>Results</strong></h3><div>The findings demonstrate several opportunities of GenAI in Software Engineering to increase productivity in development teams. However, several key barriers were also identified, that should be accounted for in successful integrations. We synthesize the results in a grounded conceptual framework for GenAI adoption in Software Engineering.</div></div><div><h3><strong>Conclusions</strong></h3><div>This study contributes to the discourse on GenAI in Software Engineering by providing a conceptual framework that aids in understanding the opportunities and challenges of GenAI. It offers practical guidelines for businesses and developers to enhance GenAI integration and lays the groundwork for future research on its impact in software development.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107751"},"PeriodicalIF":3.8,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143817125","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Understanding the product knowledge acquisition process in multiprovider software evolution scenarios: An exploratory study","authors":"Anelis Pereira-Vale,&nbsp;Tomás Vera ,&nbsp;Carlos Vásquez ,&nbsp;Daniel Perovich,&nbsp;Jocelyn Simmonds,&nbsp;Sergio F. Ochoa","doi":"10.1016/j.infsof.2025.107738","DOIUrl":"10.1016/j.infsof.2025.107738","url":null,"abstract":"<div><h3>Context</h3><div>The life cycle of a custom software system involves multiple evolution projects, where each one can be carried out by a different provider, i.e., software development company. When the provider changes, the new one should understand the structure and functionality of the product to evolve, in order to determine the scope, effort, risks and uncertainties of the next evolution project. Regardless of its relevance in practice, product comprehension has received little attention, particularly in multiprovider evolution scenarios. This limits the ability of researchers and practitioners to understand and improve this process.</div></div><div><h3>Objective</h3><div>To understand the process performed by software teams to discover and assimilate the knowledge about the structure, functionality, and quality aspects of a custom system developed by another team.</div></div><div><h3>Method</h3><div>We conducted an exploratory study that answered four research questions considering two particular software evolution scenarios. The study involved interviews with 19 project leaders from Chilean and Brazilian small and medium-sized enterprises, who regularly participate in this type of project.</div></div><div><h3>Results</h3><div>For each evolution scenario, we identified: (1) the underlying structure and dynamic of the product knowledge acquisition process, (2) the knowledge gathering activities and software artifacts used by the development teams, and (3) how effective these teams were when performing these activities. The study also identifies causes of low effectiveness of teams in this process and promising research avenues to address them.</div></div><div><h3>Conclusion</h3><div>The product knowledge acquisition process is usually informal and has low cost-effectiveness. Provider switching results in an important loss of information and knowledge about the structure and functionality of the product, which limits comprehension and evolution of the system. The results show a clear need to change the status-quo in the customer-provider relationship in the study scenarios, and open several research opportunities to improve this process.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107738"},"PeriodicalIF":3.8,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785843","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Uncertainty propagation from sensor data to deep learning models in autonomous driving","authors":"Yifan Wang ,&nbsp;Tiexin Wang ,&nbsp;Tao Yue","doi":"10.1016/j.infsof.2025.107735","DOIUrl":"10.1016/j.infsof.2025.107735","url":null,"abstract":"<div><h3>Context:</h3><div>Deep learning has been widely used in Autonomous Driving Systems (ADS). Though significant progress has been made regarding their efficiency and accuracy, uncertainty remains a critical factor affecting ADS safety. Such uncertainties are often due to environmental noise and/or imperfect algorithm structures. Studies on uncertainty quantification mostly focus on single classification tasks and overlook how uncertainties propagate from the perception to downstream decision-making, studying of which is critical, as the interplay between perception and decision-making can significantly impact the overall safety of ADS.</div></div><div><h3>Objectives:</h3><div>We quantify and understand the uncertainty propagation from sensor data to deep learning models, as well as its impact on ADS safety.</div></div><div><h3>Methods:</h3><div>We present an empirical study that quantifies both aleatoric and epistemic uncertainties and assesses how such uncertainties propagate and impact ADS safety under various sensor noise conditions. We also investigate the suitability of two epistemic uncertainty quantification methods (i.e., MC Dropout and Deep Ensembles) to ADS tasks and their cost-effectiveness in selecting highly-uncertain samples.</div></div><div><h3>Results:</h3><div>Results show that increased noise can significantly increase uncertainty and degrade model performance, thereby compromising decision-making and potentially impacting ADS safety. Both MC Dropout and Deep Ensembles effectively measure the model’s epistemic uncertainty, with MC Dropout showing higher correlation with ADS safety, and saving time and computational costs. Moreover, there are significant differences in the highly-uncertain samples they identified.</div></div><div><h3>Conclusion:</h3><div>Our results show the importance of considering uncertainty propagation to ensure the ADS safety. Compared to Deep Ensembles, MC Dropout’s efficiency makes it a more suitable choice in the context of ADS.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"183 ","pages":"Article 107735"},"PeriodicalIF":3.8,"publicationDate":"2025-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}