IET Software最新文献

{"title":"Fusing General and Domain-Specific Transformers for Fine-Grained Emotion Detection","authors":"Binh Vu,&nbsp;Harshita Jamadade,&nbsp;Swati Chandna,&nbsp;Mehrdad Jalali","doi":"10.1049/sfw2/5612036","DOIUrl":"https://doi.org/10.1049/sfw2/5612036","url":null,"abstract":"<p>The proliferation of user-generated content has transitioned affective computing toward fine-grained emotion recognition. Standard pretrained transformers encounter representational bottlenecks when processing the informal and implicit language characteristic of online discourse, a challenge severely exacerbated by extreme class imbalance. This paper addresses these gaps by proposing a dual-stream hybrid neural architecture for multilabel emotion classification on the GoEmotions dataset. The framework processes text through parallel pathways: a general-purpose language model capturing broad semantics and a domain-oriented encoder adapted to social media vernacular. To overcome data-driven limitations under severe label sparsity, the architecture integrates affective commonsense knowledge from SenticNet via a bounded breadth–first search over concept networks and models interlabel dependencies through an adaptive sparse self-attention label embedding module regularized by an auxiliary multitask objective. Every configuration, including all baselines, is evaluated under an identical per-class threshold calibration procedure, and the principal comparisons are replicated across 10 random seeds with paired significance testing. Under this protocol, the proposed model attains a macro F1 of 0.5368 across 28 categories, exceeding the strongest calibrated single-stream baseline. Component-wise ablation attributes this gain to the knowledge integration rather than to encoder fusion, which confers no standalone benefit despite doubling parameters and inference latency. Two findings of independent methodological interest emerge. First, decision-threshold policy is a first-order confound in long-tail emotion evaluation: A single trained model moves by 0.095 macro F1 between uniform and per-class thresholding, an order of magnitude more than any architectural intervention we measure, and enough to reverse the conclusion that inverse-frequency reweighting is harmful. Second, damped positive class weighting evaluated under matched calibration is the dominant contributor to rare-class performance, worth +0.0574 macro F1 on 10 of 10 seeds and raising categories such as pride, grief, and relief from near-zero to workable F1. Cross-dataset evaluation on SemEval-2018 yields a Jaccard index of 0.613, confirming the architecture is effective on a second benchmark. The results indicate that structured affective knowledge yields measurable but modest gains and that evaluation protocol substantially conditions their interpretation.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/5612036","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149146810","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"WASS: A Weakness-Aware Subset Selection Framework for Source-Sharing-Constrained Cross-Project Defect Prediction","authors":"Haihao Zhou,&nbsp;Yongchang Ding,&nbsp;Xiao-Yuan Jing,&nbsp;Rong Peng","doi":"10.1049/sfw2/6370008","DOIUrl":"https://doi.org/10.1049/sfw2/6370008","url":null,"abstract":"<p>Cross-project defect prediction (CPDP) is a promising solution for new software projects lacking historical data. However, modern privacy/compliance constraints and limited sharing budgets often restrict how many labeled source instances can be transferred across organizations. Existing instance selection strategies typically prioritize “safe” majority instances or rely on random sampling, which can miss boundary-defining defects under very small budgets. To address this, we propose weakness-aware subset selection (WASS), a framework for source-side local curation under a source-sharing budget. WASS combines target-domain similarity, weakness-aware scoring, and class-aware quota allocation to curate a compact transferable subset of source instances. Extensive experiments on 88 directed same-family CPDP pairs with seven diverse classifiers (including transfer component analysis [TCA] + logistic regression [LR] and support vector machine [SVM]) show a clear 1%–5% Golden Window under the evaluated protocol. The experimental protocol also clarifies source-sharing accounting, repeated-run randomness, and family-level variation. Therefore, WASS is a practical, plug-and-play cold-start accelerator for source-sharing-constrained CPDP.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/6370008","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148942833","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From Fine-Tuning to Evaluation: How LLM Enables Automatic Domain Modeling","authors":"Haoran Shi,&nbsp;Gaojie Wang,&nbsp;Shijun Liu,&nbsp;Sheharyar Khan,&nbsp;Guangxu Mei,&nbsp;Li Pan","doi":"10.1049/sfw2/1690605","DOIUrl":"https://doi.org/10.1049/sfw2/1690605","url":null,"abstract":"<p>Domain modeling is widely utilized in the design, development, and maintenance processes of manufacturing scenarios. However, due to the specialized nature of domain-specific business requirements and the structural complexity of domain models, modeling remains predominantly manual or semi-manual. Artificial intelligence (AI) technologies, particularly those based on natural language processing (NLP), have made automatic modeling feasible. Nevertheless, the lack of domain-specific knowledge limits the completeness of automatic modeling achieved by such methods. To bridge this gap, we leverage large language models (LLMs) with enhanced capabilities for understanding domain-specific contexts to achieve automatic modeling for manufacturing scenarios. In this study, an automatic modeling LLM (AMLLM) is developed, enabling efficient automatic construction of complex domain models and incorporating a series of quantitative indicators and evaluation measures to assess the generated domain models’ completeness comprehensively. Supported by model evaluation, AMLLM can achieve continuous self-optimization to support more complete model generation. Finally, through comprehensive experiments guided by our research questions (RQs), we demonstrate that the proposed approach enables efficient and complete automatic domain modeling. Compared with traditional modeling approaches, our solution innovates by automatic modeling via LLMs while enhancing completeness, delivering significant efficiency and quality advantages.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/1690605","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148943744","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"RDDR-YOLO: Real-Time Lightweight Human Detection for UAV-Based Surveillance","authors":"Xiaoyang Wang,&nbsp;Guohui Zeng,&nbsp;Bo Huang,&nbsp;Huaqing Wei,&nbsp;Zhenhua Zhang,&nbsp;Gaojian Li","doi":"10.1049/sfw2/3809710","DOIUrl":"https://doi.org/10.1049/sfw2/3809710","url":null,"abstract":"<p>Driven by the rapid advancement of un-manned aerial vehicle (UAV) and deep learning technologies, UAV surveillance has emerged as a vital supplement to conventional ground monitoring systems due to its flexible deployment, wide coverage, and real-time capability. However, most state-of-the-art human action recognition models are trained on ground-level video data and struggle to adapt to the inherent characteristics of aerial imagery, including extremely small object scales, low resolution, complex background interference, and severe mutual occlusion. To address these technical bottlenecks in UAV-based human detection, this paper presents a lightweight yet high-precision detection method, dubbed RDDR-YOLO, based on YOLOv8s. Extensive experiments are conducted on two public benchmark datasets, VisDrone2019 and TinyPerson. Results show that RDDR-YOLO achieves mAP50 values of 43.9% and 36.4% on the two datasets, outperforming the YOLOv8s baseline by 7.0% and 9.6%, respectively. Its human localization precision reaches 41.2%, yielding an improvement of 14.5%. Moreover, the parameter count is reduced by 67.6%, and the inference speed reaches 70 FPS. The proposed method substantially improves the detection accuracy of small-scale human targets in UAV scenarios while maintaining competitive real-time performance. It can serve as the 2D localization module in the first stage of a two-stage human action recognition framework, effectively supporting human pose estimation and motion capture in UAV-based scenarios. The method holds significant practical potential in smart city security, emergency search and rescue, crowd management, and other related fields.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/3809710","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148784693","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Hybrid Explainable Machine Learning Framework for Fair and Transparent Loan Approval Prediction","authors":"Zohaib Ahmed Dahri,&nbsp;Fida Hussain Dahri,&nbsp;Asif Ali Laghari,&nbsp;Muhammad Javed,&nbsp;Teerath Kumar","doi":"10.1049/sfw2/5599870","DOIUrl":"https://doi.org/10.1049/sfw2/5599870","url":null,"abstract":"<p>The increasing adoption of automated systems in financial organisations demands intelligent decision-making systems that are not only accurate but also transparent, fair and explainable. However, many existing loan approval approaches primarily emphasise classification performance while providing limited integration of explainability and fairness assessment for different stakeholders. To address this gap, this paper introduces a hybrid explainable machine learning (ML) framework that combines classification performance with multi-level interpretability and fairness auditing for loan approval decisions. Using a real-world dataset of 10,000 applications, we perform extensive feature engineering, introducing novel composite variables such as movable and immovable assets to enhance the financial representativeness of applicants. We evaluate five classifiers: Decision Tree (DT), Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM) and Extreme Gradient Boosting (XGBoost). Among them, XGBoost achieved competitive results with 98.00% accuracy, 0.99 precision, 0.97 recall, 0.98 F1-score and 1.00 area under the receiver operating characteristic curve (AUC–ROC), achieving competitive performance relative to recent benchmark studies. Our framework further integrates a multi-stakeholder explainability suite: SHAP for global insights, local interpretable model-agnostic explanations (LIME) for local decision rationale, diverse counterfactual explanations (DiCE) for counterfactual guidance and AIF360 for fairness evaluation. This framework is designed to support different stakeholders, including data scientists, loan officers, applicants and regulators, by providing tailored transparency. The proposed framework demonstrates that explainability and fairness can coexist with strong classification performance, offering a more transparent, fairness-aware approach to explainable credit risk assessment.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/5599870","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148467640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FactCheck: Feasibility-Aware Long-Term Action Anticipation With Multi-Agent Collaboration","authors":"Rui Cao,&nbsp;Jiannong Cao,&nbsp;Bo Yuan,&nbsp;Zhiyuan Wen,&nbsp;Mingjin Zhang","doi":"10.1049/sfw2/7118691","DOIUrl":"https://doi.org/10.1049/sfw2/7118691","url":null,"abstract":"<p>Long-term action anticipation (LTA) aims to predict an ordered sequence of future verb-noun actions from a partially observed video. While this task serves as the foundation for embodied intelligence, anticipating physically feasible long-term actions remains a critical challenge. Existing methods, which operate in an open-loop manner, often hallucinate non-existent objects, violate object affordances, or disregard object states, as they lack explicit mechanisms to verify action feasibility against the physical environment. To address this, we propose FactCheck, a novel multi-agent collaboration framework that improves feasibility through a closed-loop “Observe-Plan-Verify” mechanism. FactCheck decomposes the complex LTA task into specialized roles: an Observer that recognizes historical actions from video observations and constructs a dual-form structured memory, comprising a History Action Abstract that captures high-level human intentions and environmental status, and a History Action Graph that encodes object states and temporal dependencies; a Planner that generates draft future actions conditioned on both low-level historical actions and high-level History Action Abstract; and a Verifier that rigorously validates the draft against the History Action Graph and refines infeasible actions. Extensive experiments on the EPIC-Kitchens-55 and EGTEA Gaze+ benchmarks demonstrate that FactCheck consistently outperforms state-of-the-art (SoTA) methods. Our work establishes a new paradigm for feasibility-aware LTA, effectively closing the loop of action recognition, action prediction, and action verification.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/7118691","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148467036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Hybrid Psychometric-MCDM Framework for Optimized Human-Centric Decisions in Software Engineering","authors":"Abdulaziz Alhubaishy","doi":"10.1049/sfw2/9921631","DOIUrl":"https://doi.org/10.1049/sfw2/9921631","url":null,"abstract":"<p>Decision-making in software development involves balancing multiple technical and organizational factors. While multicriteria decision-making (MCDM) methods offer systematic approaches to such problems, they typically do not account for practitioners’ emotional states, which may influence judgments. This study examines how affective states impact decision-making and proposes a framework integrating psychometric evaluation with MCDM techniques. We present a framework combining MCDM, specifically the best–worst method (BWM) for problem formulation, with an affective scale, specifically the International Positive and Negative Affect Schedule - Short Form (I-PANAS-SF), to estimate the individual’s affective state and mitigate the potential negative influence it might have on their decision. An experiment was conducted to observe how the valence of the affective state (whether positive or negative) may impact the importance assigned to decision criteria. Evaluations were filtered to include inputs from participants exhibiting positive affect, and the resulting prioritization outcomes were compared with an external benchmark dataset. The results indicate that the framework produced different aggregated decisions depending on affective valence, with positive-affect filtering yielding distinct prioritization patterns. Whether these differences improve decision quality requires direct comparison with affect-agnostic baselines. The exploratory comparison with the benchmark dataset revealed variations in task valuation that may be associated with affective states. These findings demonstrate the feasibility of incorporating psychometric assessment into MCDM-based software engineering (SE) decisions while highlighting the need for further validation using larger and more diverse samples, as well as careful consideration of ethical and practical deployment challenges.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/9921631","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148466718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Empirical Process Model for Agile Practitioners Developing Healthcare Software","authors":"Yazidu Buba Salihu,&nbsp;Julian Michael Bass,&nbsp;Gloria Ejehiohen Iyawa","doi":"10.1049/sfw2/6862755","DOIUrl":"https://doi.org/10.1049/sfw2/6862755","url":null,"abstract":"<p>Agile philosophy emphasises flexibility, adaptability and human-centred software development. However, in medical software development, integrating regulatory compliance and safety-critical requirements within Agile methods introduces significant tension. Compliance and safety-critical are often treated as afterthoughts rather than integrated in a dynamic, Agile workflow, leading to costly rework and system failures. To address these challenges, this study employed a qualitative, multi-method research design spanning four phases. Two phases involved in-depth interviews with 30 Agile practitioners from the UK, Nigeria and India, recruited using snowball sampling. In the third phase, a novel practice-based process model was developed. To establish the model’s practice relevance and transferability, the model was evaluated with eight Agile practitioners in the UK and Nigeria in the fourth phase. To improve credibility, the model was further validated through a case study involving Nigerian practitioners managing a Canadian medical software project. Data from each phase were inductively analysed using techniques informed by grounded theory: open coding, memoing, constant comparison and theoretical saturation. The study identified 39 Agile practices used by practitioners for medical software development. The findings reveal three novel artefacts, the compliance backlog, safety clinical backlog and misuse backlog, and two novel ceremonies: internal clinical demoing and the compliance review. In addition, three specialised roles emerged: the medical scrum master, clinical tester and regulatory expert. These practices were not theorised a priori but emerged inductively from practitioners’ interviews. Practitioners reported that regulatory compliance and safety outcomes improve when regulatory and safety-critical requirements are embedded dynamically within Agile workflows rather than treated as external constraints. This research contributes a novel, empirically grounded taxonomy of Agile roles, ceremonies and artefacts that goes beyond merely proposing another process design; it presents an operationalised Agile workflow shaped by the real experiences of Agile practitioners in regulated medical software environments. Reflecting how compliance and safety activities are coordinated in practice, the study demonstrates the embedding of ’compliance and safety-by-design’ within Agile software development. We recommend using the model, as it includes novel practices for reconciling contextual influences, including weak regulatory implementation and siloed development practices.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/6862755","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148466719","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Blockchain-Enabled AI for Smart Head Counting and Emergency Mustering System","authors":"Muhammad Saad,&nbsp;Maaz Bin Ahmad,&nbsp;Muhammad Asif,&nbsp;Syed Mubashir Ali,&nbsp;Monia Hamdi,&nbsp;Asif Ali Zain Ul Abdin,&nbsp;Nadeem Sarwar","doi":"10.1049/sfw2/3472254","DOIUrl":"10.1049/sfw2/3472254","url":null,"abstract":"<p>This research provides a comprehensive framework for the digitalization of head counting and the emergency mustering system. An end-to-end framework and solution is proposed for the developing countries by prioritizing health and safety, particularly in the industrial sector, where a large number of employees face significant risks during chaotic situations. It addresses the limitations of manual and traditional head counting during emergencies by providing real-time analytics through facial recognition, enabling the identification of accounted-for and missing personnel. The work integrates blockchain as compared to traditional methods along with the Internet of Things (IoT), Messaging Queue Telemetry Transport (MQTT), and wireless technologies to create an end-to-end framework for automating the headcount process. The study emphasizes using blockchain and IoT-enabled terminals, facilitating seamless data transmission to a blockchain-enabled network. The cross-platform software application is also developed to represent the data in the form of a dashboard on the web and mobile. The pilot project was implemented in an industrial environment to collect experimental logs and validate the robustness of this highly integrated system. The proposed system achieved 99.8% accuracy with a precision of 99.6%, a recall of 99.7%, and an F1-score of 99.65%, reducing errors from 8% to 0.2% and reducing the emergency response time by 35% as compared to traditional roll-call methods. Future research may focus on scaling the system using artificial intelligence (AI) cameras and 5G technology to enhance mobility and headcount accuracy.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/3472254","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148381129","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Radiology Report Generation via Self-Critical Multisource Knowledge Reasoning With Large Language Models","authors":"Sharofiddin Allaberdiev,&nbsp;Xiaojun Chen,&nbsp;Dilmurat Mamatov,&nbsp;Akmalbek Abdusalomov,&nbsp;Utkirjon Ubaydullaev,&nbsp;Asif Khan","doi":"10.1049/sfw2/5529107","DOIUrl":"https://doi.org/10.1049/sfw2/5529107","url":null,"abstract":"<p>Automated radiology report generation (RRG) aims to reduce radiologist workload while improving reporting consistency and clinical reliability. Despite its importance for diagnosis, manual reporting remains time-consuming and error-prone. Existing encoder–decoder models often struggle to align visual and textual representations, resulting in incomplete or inconsistent reports. Large language models (LLMs) provide strong linguistic fluency for automated RRG, but most existing approaches still follow a linear generation paradigm and lack explicit reasoning, self-evaluation, and structured clinical grounding. To address these limitations, we propose SMK–RRG, a self-critical multisource Knowledge framework for RRG. The proposed framework integrates structured clinical knowledge, pathology-aware visual evidence, and iterative self-refinement within a unified LLM-driven architecture. It is implemented through self-critical thought (SCT) prompting, which performs multistage candidate report generation, LLM-based self-critique, and CheXbert-guided clinical selection among candidate reports. By enabling critique-guided prompt evolution across refinement stages, SCT moves beyond conventional chain-based prompting and supports nonlinear, clinically grounded reasoning. Experiments on public chest X-ray benchmarks demonstrate that SMK–RRG produces more accurate, interpretable, and clinically consistent reports, highlighting the effectiveness of self-critique and structured information integration for reliable clinical reports.</p>","PeriodicalId":50378,"journal":{"name":"IET Software","volume":"2026 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/sfw2/5529107","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148324561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}