arXiv subjects

Guoping Rong

Publications and source records attributed to Guoping Rong.

17 recordsLinked to original sources

arXiv2026

OpenCodeReview: Determinism over Non-Determinism for Cost-Effective Agent-Based Code Review

LLM-based code review agents promise scalable, always-on review, yet current systems suffer from two intertwined weaknesses: (1) non-determinism--unbounded tool use makes review outcomes unstable, and (2) context locality--the reviewer's access remains bounded to the diff, capping discoverable issue depth. Both give rise to three challenges: misaligned context retrieval, a coherence-efficiency trade-off in multi-file pull requests, and hallucinated comments that erode trust. To address these, we introduce OpenCodeReview, built on deterministic engineering for uncertain agents: rather than granting maximal freedom, we inject determinism at three deliberate pipeline points. Rule-Guided Dispatch uses a multi-layer rule system to deterministically select files and review criteria, eliminating variability of agent-driven triage. Grounded File Review replaces free-form exploration with a curated tool set exposed through a ReAct loop, while file-level parallel SubAgents balance context coherence against efficiency and recover cross-file dependencies on demand. Independent Reflection introduces a falsification-first filter under an asymmetric information boundary--the reflector sees only the diff, not the agent's tool-augmented exploration--removing hallucinated comments without self-reinforcing bias, improving precision while preserving recall. On AACR-Bench (200 real-world PRs, 10 languages, 1,505 expert-verified comments), OpenCodeReview outperforms mainstream coding agents (e.g., Claude Code and Codex) across six LLM backends, achieving up to 2.17x higher SEM-F1 (25.10% vs. 11.57%) while consuming 5-15x fewer tokens. We open-source OpenCodeReview at https://github.com/alibaba/open-code-review.

cs.SE↗

arXiv2026

Improving LLM-Based Go Code Review through Issue-List Generation and Context Augmentation

LLMs have shown strong potential for automating code review, yet their practical utility depends heavily on the design of generation and context strategies. In this paper, we investigate how to improve LLM-based code review through generation strategy and contextual augmentation. We first propose an issue-list review paradigm, in which LLMs enumerate all potential issues rather than reporting only the single most important one (i.e., primary-issue review). We then systematically compare three types of code context augmentation -- neighboring, LSP-based semantics, and IR-based similar co-change context -- and study how they influence issue discovery. Finally, we integrate candidates from no-context and context-enhanced generation to improve review coverage, and introduce refinement-guided pruning to keep the candidate list at a practical size. We evaluate our approach on 1,438 Go review instances using downstream code refinement as the main metric, i.e., how often the candidate list contains at least one comment inducing the same code change as the final human revision. For comparison, we evaluate comments by CodeReviewer, a model trained specifically for review comment generation, as well as ground-truth human review comments (as a practical upper bound), under the same refinement-based evaluation. The results show that our best configuration, combining issue-list review, neighboring and similar co-change context, and candidate integration, reaches 28.00% refinement exact match, a statistically significant gain of +10.85 percentage points over primary-issue review without any additional context (17.15%), substantially outperforming CodeReviewer (15.02%) and approaching the human-oracle ceiling of 36.09%. Our refinement-guided pruning reduces the average candidate count from 7.2 to 3.1 at top-5 while retaining nearly the full benefit, making the candidate list easier to inspect.

cs.SE↗

arXiv2026

ConceptRM: The Quest to Mitigate Alert Fatigue through Consensus-Based Purity-Driven Data Cleaning for Reflection Modelling

In many applications involving intelligent agents, the overwhelming volume of alerts (mostly false) generated by the agents may desensitize users and cause them to overlook critical issues, leading to the so-called ''alert fatigue''. A common strategy is to train a reflection model as a filter to intercept false alerts with labelled data collected from user verification feedback. However, a key challenge is the noisy nature of such data as it is often collected in production environments. As cleaning noise via manual annotation incurs high costs, this paper proposes a novel method ConceptRM for constructing a high-quality corpus to train a reflection model capable of effectively intercepting false alerts. With only a small amount of expert annotations as anchors, ConceptRM creates perturbed datasets with varying noise ratios and utilizes co-teaching to train multiple distinct models for collaborative learning. By analyzing the consensus decisions of these models, it effectively identifies reliable negative samples from a noisy dataset. Experimental results demonstrate that ConceptRM significantly enhances the interception of false alerts with minimal annotation cost, outperforming several state-of-the-art LLM baselines by up to 53.31% on in-domain datasets and 41.67% on out-of-domain datasets.

cs.CL↗

arXiv2026

AACR-Bench: Evaluating Automatic Code Review with Holistic Repository-Level Context

High-quality evaluation benchmarks are pivotal for deploying Large Language Models (LLMs) in Automated Code Review (ACR). However, existing benchmarks suffer from two critical limitations: first, the lack of multi-language support in repository-level contexts, which restricts the generalizability of evaluation results; second, the reliance on noisy, incomplete ground truth derived from raw Pull Request (PR) comments, which constrains the scope of issue detection. To address these challenges, we introduce AACR-Bench a comprehensive benchmark that provides full cross-file context across multiple programming languages. Unlike traditional datasets, AACR-Bench employs an "AI-assisted, Expert-verified" annotation pipeline to uncover latent defects often overlooked in original PRs, resulting in a 285% increase in defect coverage. Extensive evaluations of mainstream LLMs on AACR-Bench reveal that previous assessments may have either misjudged or only partially captured model capabilities due to data limitations. Our work establishes a more rigorous standard for ACR evaluation and offers new insights on LLM based ACR, i.e., the granularity/level of context and the choice of retrieval methods significantly impact ACR performance, and this influence varies depending on the LLM, programming language, and the LLM usage paradigm e.g., whether an Agent architecture is employed. The code, data, and other artifacts of our evaluation set are available at https://github.com/alibaba/aacr-bench .

cs.SE↗

arXiv2025

Fine-Tuning LLMs to Analyze Multiple Dimensions of Code Review: A Maximum Entropy Regulated Long Chain-of-Thought Approach

Large Language Models (LLMs) have shown great potential in supporting automated code review due to their impressive capabilities in context understanding and reasoning. However, these capabilities are still limited compared to human-level cognition because they are heavily influenced by the training data. Recent research has demonstrated significantly improved performance through fine-tuning LLMs with code review data. However, compared to human reviewers who often simultaneously analyze multiple dimensions of code review to better identify issues, the full potential of these methods is hampered by the limited or vague information used to fine-tune the models. This paper contributes MelcotCR, a chain-of-thought (COT) fine-tuning approach that trains LLMs with an impressive reasoning ability to analyze multiple dimensions of code review by harnessing long COT techniques to provide rich structured information. To address context loss and reasoning logic loss issues that frequently occur when LLMs process long COT prompts, we propose a solution that combines the Maximum Entropy (ME) modeling principle with pre-defined reasoning pathways in MelcotCR to enable more effective utilization of in-context knowledge within long COT prompts while strengthening the logical tightness of the reasoning process. Empirical evaluations on our curated MelcotCR dataset and the public CodeReviewer dataset reveal that a low-parameter base model, such as 14B Qwen2.5, fine-tuned with MelcotCR can surpass state-of-the-art methods in terms of the accuracy of detecting and describing code issues, with its performance remarkably on par with that of the 671B DeepSeek-R1 model.

cs.SE↗

arXiv2025

Brevity is the Soul of Wit: Condensing Code Changes to Improve Commit Message Generation

Commit messages are valuable resources for describing why code changes are committed to repositories in version control systems (e.g., Git). They effectively help developers understand code changes and better perform software maintenance tasks. Unfortunately, developers often neglect to write high-quality commit messages in practice. Therefore, a growing body of work is proposed to generate commit messages automatically. These works all demonstrated that how to organize and represent code changes is vital in generating good commit messages, including the use of fine-grained graphs or embeddings to better represent code changes. In this study, we choose an alternative way to condense code changes before generation, i.e., proposing brief yet concise text templates consisting of the following three parts: (1) summarized code changes, (2) elicited comments, and (3) emphasized code identifiers. Specifically, we first condense code changes by using our proposed templates with the help of a heuristic-based tool named ChangeScribe, and then fine-tune CodeLlama-7B on the pairs of our proposed templates and corresponding commit messages. Our proposed templates better utilize pre-trained language models, while being naturally brief and readable to complement generated commit messages for developers. Our evaluation based on a widely used dataset showed that our approach can outperform six baselines in terms of BLEU-Norm, METEOR, and ROUGE-L, with average improvements of 51.7%, 78.7%, and 62.5%, respectively. The ablation study and human evaluation also provide further insights into the effectiveness of our approach.

cs.SE↗

arXiv2025

Does AI Code Review Lead to Code Changes? A Case Study of GitHub Actions

AI-based code review tools automatically review and comment on pull requests to improve code quality. Despite their growing presence, little is known about their actual impact. We present a large-scale empirical study of 16 popular AI-based code review actions for GitHub workflows, analyzing more than 22,000 review comments in 178 repositories. We investigate (1) how these tools are adopted and configured, (2) whether their comments lead to code changes, and (3) which factors influence their effectiveness. We develop a two-stage LLM-assisted framework to determine whether review comments are addressed, and use interpretable machine learning to identify influencing factors. Our findings show that, while adoption is growing, effectiveness varies widely. Comments that are concise, contain code snippets, and are manually triggered, particularly those from hunk-level review tools, are more likely to result in code changes. These results highlight the importance of careful tool design and suggest directions for improving AI-based code review systems.

cs.SE↗

arXiv2025

SynthCoder: A Synthetical Strategy to Tune LLMs for Code Completion

Code completion is a prominent application of Large Language Models (LLMs) in software engineering. Due to the near real-time response requirements of this task, base models with small to medium-sized parameters are typically employed, supplemented by various optimization and post-training techniques. However, these optimization methods often have trade-offs, leading to a seesaw effect where performance improvements on certain datasets or metrics are accompanied by degradations on others -- sometimes even falling below the baseline model's performance. This paper proposes SynthCoder, a model that integrates leading industry practices to achieve state-of-the-art performance on the Fill-in-the-Middle (FIM) code completion task. In specific, we first construct a diverse dataset by combining Abstract Syntax Tree (AST) node extraction with heuristics that simulate developer behavior. Then we enrich our training corpus with cross-file contextual information using the BM25 algorithm and call graphs, enhancing the model's ability to perform code completion in both file-level and repository-level scenarios. As the last step, we employ a two-stage training process using the Seed-Coder-8B-Base as the base model. First, we fine-tune the model using Curriculum Learning technology. Following this, we perform alignment using Direct Preference Optimization (DPO) with preference pairs generated through Rejection Sampling. Experimental results demonstrate that our final model excels on mainstream repository-level code completion benchmarks, including aiXcoder, ExecRepoBench, CrossCodeEval, and CoLT. Furthermore, our carefully curated training set effectively mitigates the model's tendency to just repeat existing code, a common issue existing in various code completion models.

cs.SE↗

arXiv2024

Distilling Desired Comments for Enhanced Code Review with Large Language Models

There has been a growing interest in using Large Language Models (LLMs) for code review thanks to their proven proficiency in code comprehension. The primary objective of most review scenarios is to generate desired review comments (DRCs) that explicitly identify issues to trigger code fixes. However, existing LLM-based solutions are not so effective in generating DRCs for various reasons such as hallucination. To enhance their code review ability, they need to be fine-tuned with a customized dataset that is ideally full of DRCs. Nevertheless, such a dataset is not yet available, while manual annotation of DRCs is too laborious to be practical. In this paper, we propose a dataset distillation method, Desiview, which can automatically construct a distilled dataset by identifying DRCs from a code review dataset. Experiments on the CodeReviewer dataset comprising more than 150K review entries show that Desiview achieves an impressive performance of 88.93%, 80.37%, 86.67%, and 84.44% in terms of Precision, Recall, Accuracy, and F1, respectively, surpassing state-of-the-art methods. To validate the effect of such a distilled dataset on enhancing LLMs' code review ability, we first fine-tune the latest LLaMA series (i.e., LLaMA 3 and LLaMA 3.1) to build model Desiview4FT. We then enhance the model training effect through KTO alignment by feeding those review comments identified as non-DRCs to the LLMs, resulting in model Desiview4FA. Verification results indicate that Desiview4FA slightly outperforms Desiview4FT, while both models have significantly improved against the base models in terms of generating DRCs. Human evaluation confirms that both models identify issues more accurately and tend to generate review comments that better describe the issues contained in the code than the base LLMs do.

cs.SE↗

arXiv2024

AUCAD: Automated Construction of Alignment Dataset from Log-Related Issues for Enhancing LLM-based Log Generation

Log statements have become an integral part of modern software systems. Prior research efforts have focused on supporting the decisions of placing log statements, such as where/what to log. With the increasing adoption of Large Language Models (LLMs) for code-related tasks such as code completion or generation, automated approaches for generating log statements have gained much momentum. However, the performance of these approaches still has a long way to go. This paper explores enhancing the performance of LLM-based solutions for automated log statement generation by post-training LLMs with a purpose-built dataset. Thus the primary contribution is a novel approach called AUCAD, which automatically constructs such a dataset with information extracting from log-related issues. Researchers have long noticed that a significant portion of the issues in the open-source community are related to log statements. However, distilling this portion of data requires manual efforts, which is labor-intensive and costly, rendering it impractical. Utilizing our approach, we automatically extract log-related issues from 1,537 entries of log data across 88 projects and identify 808 code snippets (i.e., methods) with retrievable source code both before and after modification of each issue (including log statements) to construct a dataset. Each entry in the dataset consists of a data pair representing high-quality and problematic log statements, respectively. With this dataset, we proceed to post-train multiple LLMs (primarily from the Llama series) for automated log statement generation. Both human and experimental evaluations indicate that these models significantly outperform existing LLM-based solutions, thereby validating the efficacy of our method for constructing a post-training dataset to enhance LLM-based log statement generation.

cs.SE↗

arXiv2024

AVIATE: Exploiting Translation Variants of Artifacts to Improve IR-based Traceability Recovery in Bilingual Software Projects

Traceability plays a vital role in facilitating various software development activities by establishing the traces between different types of artifacts (e.g., issues and commits in software repositories). Among the explorations for automated traceability recovery, the IR (Information Retrieval)-based approaches leverage textual similarity to measure the likelihood of traces between artifacts and show advantages in many scenarios. However, the globalization of software development has introduced new challenges, such as the possible multilingualism on the same concept (e.g., "ShuXing" vs. "attribute") in the artifact texts, thus significantly hampering the performance of IR-based approaches. Existing research has shown that machine translation can help address the term inconsistency in bilingual projects. However, the translation can also bring in synonymous terms that are not consistent with those in the bilingual projects (e.g., another translation of "ShuXing" as "property"). Therefore, we propose an enhancement strategy called AVIATE that exploits translation variants from different translators by utilizing the word pairs that appear simultaneously across the translation variants from different kinds artifacts (a.k.a. consensual biterms). We use these biterms to first enrich the artifact texts, and then to enhance the calculated IR values for improving IR-based traceability recovery for bilingual software projects. The experiments on 17 bilingual projects (involving English and 4 other languages) demonstrate that AVIATE significantly outperformed the IR-based approach with machine translation (the state-of-the-art in this field) with an average increase of 16.67 in Average Precision (31.43%) and 8.38 (11.22%) in Mean Average Precision, indicating its effectiveness in addressing the challenges of multilingual traceability recovery.

cs.SE↗

arXiv2024

Detecting Build Dependency Errors in Incremental Builds

Incremental and parallel builds performed by build tools such as Make are the heart of modern C/C++ software projects. Their correct and efficient execution depends on build scripts. However, build scripts are prone to errors. The most prevalent errors are missing dependencies (MDs) and redundant dependencies (RDs). The state-of-the-art methods for detecting these errors rely on clean builds (i.e., full builds of a subset of software configurations in a clean environment), which is costly and takes up to multiple hours for large-scale projects. To address these challenges, we propose a novel approach called EChecker to detect build dependency errors in the context of incremental builds. The core idea of EChecker is to automatically update actual build dependencies by inferring them from C/C++ pre-processor directives and Makefile changes from new commits, which avoids clean builds when possible. EChecker achieves higher efficiency than the methods that rely on clean builds while maintaining effectiveness. We selected 12 representative projects, with their sizes ranging from small to large, with 240 commits (20 commits for each project), based on which we evaluated the effectiveness and efficiency of EChecker. We compared the evaluation results with a state-of-the-art build dependency error detection tool. The evaluation shows that the F-1 score of EChecker improved by 0.18 over the state-of-the-art method. EChecker increases the build dependency error detection efficiency by an average of 85.14 times (with the median at 16.30 times). The results demonstrate that EChecker can support practitioners in detecting build dependency errors efficiently.

cs.SE↗

arXiv2023

TRIAD: Automated Traceability Recovery based on Biterm-enhanced Deduction of Transitive Links among Artifacts

Traceability allows stakeholders to extract and comprehend the trace links among software artifacts introduced across the software life cycle, to provide significant support for software engineering tasks. Despite its proven benefits, software traceability is challenging to recover and maintain manually. Hence, plenty of approaches for automated traceability have been proposed. Most rely on textual similarities among software artifacts, such as those based on Information Retrieval (IR). However, artifacts in different abstraction levels usually have different textual descriptions, which can greatly hinder the performance of IR-based approaches (e.g., a requirement in natural language may have a small textual similarity to a Java class). In this work, we leverage the consensual biterms and transitive relationships (i.e., inner- and outer-transitive links) based on intermediate artifacts to improve IR-based traceability recovery. We first extract and filter biterms from all source, intermediate, and target artifacts. We then use the consensual biterms from the intermediate artifacts to extend the biterms of both source and target artifacts, and finally deduce outer and inner-transitive links to adjust text similarities between source and target artifacts. We conducted a comprehensive empirical evaluation based on five systems widely used in other literature to show that our approach can outperform four state-of-the-art approaches, and how its performance is affected by different conditions of source, intermediate, and target artifacts. The results indicate that our approach can outperform baseline approaches in AP over 15% and MAP over 10% on average.

cs.SE↗

arXiv2022

Using Consensual Biterms from Text Structures of Requirements and Code to Improve IR-Based Traceability Recovery

Traceability approves trace links among software artifacts based on whether two artifacts are related by system functionalities. The traces are valuable for software development, but are difficult to obtain manually. To cope with the costly and fallible manual recovery, automated approaches are proposed to recover traces through textual similarities among software artifacts, such as those based on Information Retrieval (IR). However, the low quality & quantity of artifact texts negatively impact the calculated IR values, thus greatly hindering the performance of IR-based approaches. In this study, we propose to extract co-occurred word pairs from the text structures of both requirements and code (i.e., consensual biterms) to improve IR-based traceability recovery. We first collect a set of biterms based on the part-of-speech of requirement texts, and then filter them through the code texts. We then use these consensual biterms to both enrich the input corpus for IR techniques and enhance the calculations of IR values. A nine-system-based evaluation shows that in general, when solely used to enhance IR techniques, our approach can outperform pure IR-based approaches and another baseline by 21.9% & 21.8% in AP, and 9.3% & 7.2% in MAP, respectively. Moreover, when used to collaborate with another enhancing strategy from different perspectives, it can outperform this baseline by 5.9% in AP and 4.8% in MAP.

cs.SE↗

arXiv2022

An Industrial Experience Report on Retro-inspection

To reinforce the quality of code delivery, especially to improve future coding quality, one global Information and Communication Technology (ICT) enterprise has institutionalized a retrospective style inspection (namely retro-inspection), which is similar to Fagan inspection but differs in terms of stage, participants, etc. This paper reports an industrial case study that aims to investigate the experiences and lessons from this software practice. To this end, we collected and analyzed various empirical evidence for data triangulation. The results reflect that retro-inspection distinguishes itself from peer code review by identifying more complicated and underlying defects, providing more indicative and suggestive comments. Many experienced inspectors indicate defects together with their rationale behind and offer suggestions for correction and prevention. As a result, retro-inspection can benefit not only quality assurance (like Fagan inspection), but also internal audit, inter-division communication, and competence promotion. On the other side, we identify several lessons of retro-inspection at this stage, e.g., developers' acceptance and organizers' predicament, for next-step improvement of this practice. To be specific, some recommendations are discussed for retro-inspection, e.g., more adequate preparation and more careful publicity. This study concludes that most of the expected benefits of retro-inspection can be empirically confirmed in this enterprise and its value on the progress to continuous maturity can be recognized organization-wide. The experiences on executing this altered practice in a large enterprise provide reference value on code quality assurance to other software organizations.

cs.SE↗

arXiv2022

Modeling Review History for Reviewer Recommendation:A Hypergraph Approach

Modern code review is a critical and indispensable practice in a pull-request development paradigm that prevails in Open Source Software (OSS) development. Finding a suitable reviewer in projects with massive participants thus becomes an increasingly challenging task. Many reviewer recommendation approaches (recommenders) have been developed to support this task which apply a similar strategy, i.e. modeling the review history first then followed by predicting/recommending a reviewer based on the model. Apparently, the better the model reflects the reality in review history, the higher recommender's performance we may expect. However, one typical scenario in a pull-request development paradigm, i.e. one Pull-Request (PR) (such as a revision or addition submitted by a contributor) may have multiple reviewers and they may impact each other through publicly posted comments, has not been modeled well in existing recommenders. We adopted the hypergraph technique to model this high-order relationship (i.e. one PR with multiple reviewers herein) and developed a new recommender, namely HGRec, which is evaluated by 12 OSS projects with more than 87K PRs, 680K comments in terms of accuracy and recommendation distribution. The results indicate that HGRec outperforms the state-of-the-art recommenders on recommendation accuracy. Besides, among the top three accurate recommenders, HGRec is more likely to recommend a diversity of reviewers, which can help to relieve the core reviewers' workload congestion issue. Moreover, since HGRec is based on hypergraph, which is a natural and interpretable representation to model review history, it is easy to accommodate more types of entities and realistic relationships in modern code review scenarios. As the first attempt, this study reveals the potentials of hypergraph on advancing the pragmatic solutions for code reviewer recommendation.

cs.SE↗

arXiv2021

Exploiting the Unique Expression for Improved Sentiment Analysis in Software Engineering Text

Sentiment analysis on software engineering (SE) texts has been widely used in the SE research, such as evaluating app reviews or analyzing developers sentiments in commit messages. To better support the use of automated sentiment analysis for SE tasks, researchers built an SE-domain-specified sentiment dictionary to further improve the accuracy of the results. Unfortunately, recent work reported that current mainstream tools for sentiment analysis still cannot provide reliable results when analyzing the sentiments in SE texts. We suggest that the reason for this situation is because the way of expressing sentiments in SE texts is largely different from the way in social network or movie comments. In this paper, we propose to improve sentiment analysis in SE texts by using sentence structures, a different perspective from building a domain dictionary. Specifically, we use sentence structures to first identify whether the author is expressing her sentiment in a given clause of an SE text, and to further adjust the calculation of sentiments which are confirmed in the clause. An empirical evaluation based on four different datasets shows that our approach can outperform two dictionary-based baseline approaches, and is more generalizable compared to a learning-based baseline approach.

cs.SE↗