检索结果-内蒙古大学图书馆

Just-In-Time Detection of Outdated Comments in Software Development by Jointly Reasoning

IEEE Transactions on Consumer Electronics 2025年

作者： Tan, Jiaxi Guo, Shikai Tao, Zijian Yang, Zhiguo Li, Hui Dalian Maritime University School of Information Science and Technology Dalian China Key Laboratory of Artificial Intelligence of Dalian Dalian China Missouri State University College of Business SpringfieldMO United States

Code comments are a crucial source of software to learn various aspects of code. However, with the iterative upgrades of software, outdated code comments are increasingly prevalent. Inconsistencies in code comments can mislead developers and lead to potential errors. Due to the complexity of semantic information and interactions between source code and comments, previous studies have difficulty capturing the long-term and non-sequential dependencies in source code and the complex interaction information between code and comments. To address these challenges in outdated comment detection, we propose an approach named OutComDeter consists of the following four components, i.e., the data processing component, the jointly reasoning component, the feature extraction component and the outdated comment detection component. Firstly, the data processing component transforms source code and comments into edit sequences and comment sequences. Subsequently, to address the limitations of existing methods in capturing the interactive information between code and comments, the joint reasoning component employs two encoders with co-matching attention network to incorporate the joint relational information between code and comments into their respective feature representations. This joint relational information thereby contributes to the detection of outdated code comments. Furthermore, to mitigate the interference caused by the lack of long-term and non-sequential dependencies in source code for the task of outdated code comment detection, the feature extraction component leverages multi-head attention layers and dot-product attention layers to capture these dependencies, forming the final feature representations for both code and comments. Finally, the outdated comment detection will detect the outdated comments through a non-linear transformation. The experimental results indicate that OutComDeter outperforms the state-of-the-art methods in detecting outdated comments, as evidenc

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

How to leverage large language models for automatic ICD coding

引用

Computers in Biology and Medicine 2025年 189卷 109971-109971页

作者： Yoo, Youngju Kim, Sewon KAIST Yuseong-gu Daejeon Korea Republic of NAVER Digital Healthcare Lab Seongnam-si Gyeonggi-do Korea Republic of NAVER Cloud Seongnam-si Gyeonggi-do Korea Republic of

ICD coding, which involves assigning appropriate ICD codes to clinical notes, is essential for healthcare tasks such as health expense claims, insurance claims, and disease research. Manual ICD coding is time-consuming and prone to errors, increasing the need for automation. However, clinical notes often contain non-grammatical expressions, abbreviations, professional terms, and synonyms, making them notably noisy compared to general documents. Additionally, ICD coding faces challenges such as a broad label space and the long-tail problem, making automatic ICD coding highly challenging. Large Language Models (LLMs) have shown great potential in code extraction tasks due to their exceptional natural language understanding and information extraction capabilities. However, the unique characteristics of clinical records and ICD codes necessitate fine-tuning LLMs for optimal performance in ICD coding. In this study, we propose a novel fine-tuning framework for LLMs aimed at automatic ICD coding. Our framework introduces additional elements, including a label attention mechanism, note-relevant knowledge injection based on medical expressions, and knowledge-driven sampling to address the input token limitations of LLMs. Experiments on the MIMIC-III-50 dataset show that our framework outperforms vanilla fine-tuning in both micro and macro accuracy and F1 scores, with particularly significant improvements observed in encoder–decoder models. © 2025 Elsevier Ltd

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

Optimized noise-resilient surface code teleportation interfaces

arXiv

引用

arXiv 2025年

作者： Shalby, Mohamed A. Wang, Renyu Sedov, Denis Pryadko, Leonid P. Department of Physics and Astronomy University of California Riverside92521 United States Institute for Theoretical Physics III University of Stuttgart Stuttgart70550 Germany

Connecting two surface-code patches may require significantly higher noise at the interface. We show, via circuit-level simulations under a depolarizing noise model with idle errors, that surface codes remain fault tolerant despite substantially elevated interface error rates. Specifically, we compare three strategies—direct noisy links, gate teleportation, and a CAT-state gadget—for both rotated and unrotated surface codes, and demonstrate that careful design can mitigate hook errors in each case so that the full code distance is preserved for both X and Z. Although these methods differ in space and time overhead and performance, each offers a viable route to modular surface-code architectures. Our results, obtained with Stim and PyMatching, confirm that high-noise interfaces can be integrated fault-tolerantly without compromising the code’s essential properties, indicating that fault-tolerant scaling of error-corrected modular devices is within reach with current technology. Copyright © 2025, The Authors. All rights reserved.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

Personalized Code Readability Assessment: Are We There Yet?

arXiv

引用

arXiv 2025年

作者： Vitale, Antonio Guglielmi, Emanuela Oliveto, Rocco Scalabrino, Simone DEVISER University of Molise Italy Politecnico di Torino Italy

Unreadable code could be a breeding ground for errors. Thus, previous work defined approaches based on machine learning to automatically assess code readability that can warn developers when some code artifacts (e.g., classes) become unreadable. Given datasets of code snippets manually evaluated by several developers in terms of their perceived readability, such approaches (i) establish a snippet-level ground truth, and (ii) train a binary (readable/unreadable) or a ternary (readable/neutral/unreadable) code readability classifier. Given this procedure, all existing approaches neglect the subjectiveness of code readability, i.e., the possible different developer-specific nuances in the code readability perception. In this paper, we aim to understand to what extent it is possible to assess code readability as subjectively perceived by developers through a personalized code readability assessment approach. This problem is significantly more challenging than the snippet-level classification problem: We assume that, in a realistic scenario, a given developer is keen to provide only a few code readability evaluations, thus less data is available. For this reason, we adopt an LLM with few-shot learning to achieve our goal. Our results, however, show that such an approach achieves worse results than a state-of-the-art feature-based model that is trained to work at the snippet-level. We tried to understand why this happens by looking more closely at the quality of the available code readability datasets and assessed the consistency of the inter-developer evaluations. We observed that up to a third of the evaluations are self-contradictory. Our negative results call for new and more reliable code readability datasets. © 2025, CC BY.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

AIME: AI SYSTEM OPTIMIZATION VIA MULTIPLE LLM EVALUATORS

arXiv

引用

arXiv 2024年

作者： Patel, Bhrij Chakraborty, Souradip Suttle, Wesley A. Wang, Mengdi Bedi, Amrit Singh Manocha, Dinesh University of Maryland College Park United States U.S. Army Research Laboratory MD United States Princeton University United States University of Central Florida United States

Text-based AI system optimization typically involves a feedback loop scheme where a single LLM generates an evaluation in natural language of the current output to improve the next iteration’s output. However, in this work, we empirically demonstrate that for a practical and complex task (code generation) with multiple criteria to evaluate, utilizing only one LLM evaluator tends to let errors in generated code go undetected, thus leading to incorrect evaluations and ultimately suboptimal test case performance. Motivated by this failure case, we assume there exists an optimal evaluation policy that samples an evaluation between response and ground truth. We then theoretically prove that a linear combination of multiple evaluators can approximate this optimal policy. From this insight, we propose AI system optimization via Multiple LLM Evaluators (AIME). AIME is an evaluation protocol that utilizes multiple LLMs that each independently generate an evaluation on separate criteria and then combine them via concatenation. We provide an extensive empirical study showing AIME outperforming baseline methods in code generation tasks, with up to 62% higher error detection rate and up to 16% higher success rate than a single LLM evaluation protocol on LeetCodeHard and HumanEval datasets. We also show that the selection of the number of evaluators and which criteria to utilize is non-trivial as it can impact pact success rate by up to 12%. Copyright © 2024, The Authors. All rights reserved.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

XOXO: Stealthy Cross-Origin Context Poisoning Attacks against AI coding Assistants

arXiv

引用

arXiv 2025年

作者： Štorek, Adam Gupta, Mukur Bhatt, Noopur Gupta, Aditya Kim, Janie Srivastava, Prashast Jana, Suman Columbia University United States Stanford University United States

AI coding assistants are widely used for tasks like code generation, bug detection, and comprehension. These tools now require large and complex contexts, automatically sourced from various origins—across files, projects, and contributors—forming part of the prompt fed to underlying LLMs. This automatic context-gathering introduces new vulnerabilities, allowing attackers to subtly poison input to compromise the assistant’s outputs, potentially generating vulnerable code, overlooking flaws, or introducing critical errors. We propose a novel attack, Cross-Origin Context Poisoning (XOXO), that is particularly challenging to detect as it relies on adversarial code modifications that are semantically equivalent. Traditional program analysis techniques struggle to identify these correlations since the semantics of the code remain correct, making it appear legitimate. This allows attackers to manipulate code assistants into producing incorrect outputs, including vulnerabilities or backdoors, while shifting the blame to the victim developer or tester. We introduce a novel, task-agnostic black-box attack algorithm GCGS that systematically searches the transformation space using a Cayley Graph, achieving an 83.09% attack success rate on average across five tasks and eleven models, including GPT-4o and Claude 3.5 Sonnet v2 used by many popular AI coding assistants. Furthermore, existing defenses, including adversarial fine-tuning, are ineffective against our attack, underscoring the need for new security measures in LLM-powered coding tools. © 2025, CC BY.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

Dual-Domain Exponent of Maximum Mutual Information Decoding

arXiv

引用

arXiv 2025年

作者： Moeini, AmirPouya Fàbregas, Albert Guillén i University of Cambridge United Kingdom Universitat Politècnica de Catalunya Spain

This paper provides a dual domain derivation of the error exponent of maximum mutual information (MMI) decoding with constant composition codes, showing it coincides with that of maximum likelihood decoding for discrete memoryless channels. The analysis is further extended to joint source-channel coding, demonstrating that the generalized MMI decoder achieves the same random coding error exponent as the maximum a posteriori decoder. © 2025, CC BY-NC-SA.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

Learning to Generate Unit Tests for Automated Debugging

arXiv

引用

arXiv 2025年

作者： Prasad, Archiki Stengel-Eskin, Elias Chen, Justin Chih-Yao Khan, Zaid Bansal, Mohit Department of Computer Science UNC Chapel Hill United States

Unit tests (UTs) play an instrumental role in assessing code correctness as well as providing feedback to a large language model (LLM) as it iteratively debugs faulty code, motivating automated test generation. However, we uncover a trade-off between generating unit test inputs that reveal errors when given a faulty code and correctly predicting the unit test output without access to the gold solution. To address this trade-off, we propose UTGEN, which teaches LLMs to generate unit test inputs that reveal errors along with their correct expected outputs based on task descriptions and candidate code. We integrate UTGEN into UTDEBUG, a robust debugging pipeline that uses generated tests to help LLMs debug effectively. Since model-generated tests can provide noisy signals (e.g., from incorrectly predicted outputs), UTDEBUG (i) scales UTGEN via test-time compute to improve UT output prediction, and (ii) validates and back-tracks edits based on multiple generated UTs to avoid overfitting. We show that UTGEN outperforms UT generation baselines by 7.59% based on a metric measuring the presence of both error-revealing UT inputs and correct UT outputs. When used with UTDEBUG, we find that feedback from UTGEN’s unit tests improves pass@1 accuracy of Qwen-2.5 7B on HumanEvalFix and our own harder debugging split of MBPP+ by over 3% and 12.35% (respectively) over other LLM-based UT generation baselines. Copyright © 2025, The Authors. All rights reserved.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

On low-power error-correcting cooling codes with large distances

arXiv

引用

arXiv 2024年

作者： Zhao, Yuhao Zhang, Xiande School of Mathematical Sciences University of Science and Technology of China Anhui Hefei230026 China Hefei National Laboratory University of Science and Technology of China Hefei230088 China

A low-power error-correcting cooling (LPECC) code was introduced as a coding scheme for commu- nication over a bus by Chee et al. to control the peak temperature, the average power consumption of on-chip buses, and error-correction for the transmitted information, simultaneously. Specifically, an (n, t,w, e)-LPECC code is a coding scheme over n wires that avoids state transitions on the t hottest wires and allows at most w state transitions in each transmission, and can correct up to e transmission errors. In this paper, we study the maximum possible size of an (n, t,w, e)-LPECC code, denoted by C(n, t,w, e). When w = e + 2 is large, we establish a general upper bound C(n, t,w,w - 2) ≤ ⌊(n+1/2 )/(w+t/2 )⌋;when w = e+ 2 = 3, we prove C(n, t, 3, 1) ≤ ⌊ n(n+1) 6(t+1) ⌋. Both bounds are tight for large n satisfying some divis- ibility conditions. Previously, tight bounds were known only for w = e + 2 = 3, 4 and t ≤ 2. In general, when w = e+d is large for a constant d, we determine the asymptotic value of C(n, t,w,w-d) - (n/d)/(w+t/d) as n goes to infinity, which can be extended to q-ary codes. Copyright © 2024, The Authors. All rights reserved.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

Code Review Automation Via Multi-task Federated LLM - An Empirical Study

arXiv

引用

arXiv 2024年

作者： Kumar, Jahnavi Chimalakonda, Sridhar Research in Intelligent Software & Human Analytics Lab Department of Computer Science and Engineering Indian Institute of Technology Tirupati India

Code review is a crucial process before deploying code to production, as it validates the code, provides suggestions for improvements, and identifies errors such as missed edge cases. In projects with regular production releases, the effort required for peer code-reviews remains high. Consequently, there has been significant interest from software engineering (SE) researchers in automating the code review process. Previous research on code review automation has typically approached the task as three independent sub-tasks: review necessity prediction, review comment generation, and code refinement. Our study attempts to (i) leverage the relationships between the sub-tasks of code review automation, by developing a multi-task model that addresses all tasks in an integrated manner, and (ii) increase model robustness on unseen data via collaborative large language model (LLM) modeling, while retaining the proprietary nature of code, by using federated learning (FL). The study explores five simple techniques for multi-task training, including two sequential methods, one parallel method, and two cumulative methods. The results indicate that sequentially training a federated LLM (FedLLM) for our code review multi-task use case is less efficient in terms of time, computation, and performance metrics, compared to training separate models for each task. Because sequential training demonstrates catastrophic forgetting, alternatively cumulative fine-tuning for multi-task training performs better than training models for individual tasks. This study highlights the need for research focused on effective fine-tuning of multi-task FedLLMs for SE tasks. © 2024, CC BY.

关键词： coding errors

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：