A transformer-based classification framework that fine-tunes CodeBERT on a balanced dataset of 25,000+ bug-report segments drawn from four Apache projects is proposed and CodeBERT is the best-performing transformer in this comparison.
Abstract
Maintaining code quality during software maintenance is a persistent challenge because bug-fix activities frequently introduce new code smells, accelerating long-term technical debt. Although bug-report classification has been applied to severity prediction and developer triaging, its use for proactively distinguishing Code-Quality-Impacting Bugs (CQIBs) from Non-Code-Quality-Impacting Bugs (NQIBs) remains underexplored. This paper proposes a transformer-based classification framework that fine-tunes CodeBERT on a balanced dataset of 25,000+ bug-report segments drawn from four Apache projects (Camel, CloudStack, Geode, and HBase). We evaluate two training strategies sequential transfer learning and individual fine-tuning and compare both against the published CNN-based baselines and against BERT, RoBERTa, and DeBERTa. Under 5-fold cross-validation, individual CodeBERT fine-tuning achieves weighted-F1 scores of 0.924, 0.866, 0.858, and 0.845 on Camel, CloudStack, Geode, and HBase, respectively, for an average of 0.873. Sequential transfer learning reaches a mean accuracy of 89.8% and a peak accuracy of 92.4% on Camel. CodeBERT is the best-performing transformer in this comparison, exceeding the average weighted-F1 of BERT (0.790), RoBERTa (0.753), and DeBERTa (0.742). Rigorous preprocessing, class balancing within the training folds, and stratified evaluation contribute substantially to these results.
Bug localization is the task of automatically identifying the source files responsible for a reported defect. It is a critical step in software maintenance that accelerates defect resolution. Information retrieval (IR) methods are simple and effective at exploiting historical signals such as bug-fixing recency and frequency, but they struggle to bridge the lexical gap between natural-language bug reports and programming-language identifiers. Recent work increasingly leverages pre-trained language models (PLMs) for code to close this gap. However, current PLM-based approaches still rely on a single code encoder that ignores program structure and aggregates function-level signals into file-level representations via uniform pooling. We propose a dual-encoder bug localization (DuaLoc) framework that jointly addresses these limitations. DuaLoc combines two pre-trained language models: UniXcoder for the semantic understanding of source code and GraphCodeBERT for awareness of data-flow structure. Both encoders are fine-tuned with a contrastive objective that shapes the embedding space around the localization task. A bug-report-conditioned attention mechanism then aggregates function embeddings into query-dependent file representations. The resulting neural similarity scores are then fused with classical IR features in a learning-to-rank model. DuaLoc outperforms representative classical and PLM-based baselines across most evaluation settings on a widely used benchmark of six open-source Java projects.
By combining semantic understanding, graph learning, similarity retrieval, and explainable artificial intelligence, the proposed system improves the efficiency, accuracy, and interpretability of automated bug triaging, making it suitable for large-scale software maintenance environments.
Pogiri Pujitha, Gangadhar Doma· International Journal for Re...· 0 citations
This paper evaluates various machine learning approaches on Eclipse dataset including using transformer based architectures with frozen weights as embedding with a fine-tuned neural network classification head (MLP) and experimented with Retrieval Augmented Generation (RAG).
Márk Lajkó, Balázs Nagy, László Vidács· International Conference on...· 0 citations
This research proposes the first end-to-end multi-task architecture that jointly detects Python syntax errors and localizes their exact line position by adaptively fusing CodeBERT’s semantic embeddings with handcrafted syntactic features via a lightweight gating mechanism, without relying on Abstract Syntax Trees.
Rozali Ilham, Bahtiar Imran, Hasan Basri et al.· 0 citations
This study investigates multiple experimental configurations based on CodeBERT, GraphCodeBERT, and CodeT5, including direct transformer-based classification, sliding window processing for long code sequences, and hybrid approaches combining transformer embeddings with classical machine learning classifiers.
It is suggested that process-oriented metrics, particularly those related to code testing and development history, capture defect patterns more effectively per feature than static code structure metrics, offering practical guidance for software quality assurance.
Ioana-Gabriela Chelaru, G. Czibula, Zuzsanna Oneţ-Marian et al.· Acta Universitatis Sapientia...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
