arxiv: 2604.04045 · v1 · submitted 2026-04-05 · 💻 cs.SE

Recognition: no theorem link

SmartPatchLinker: An Open-Source Tool to Linked Changes Detection for Code Review

Islem Khemissi , Moataz Chouchen , Dong Wang , Raula Gaikovina Kula

Authors on Pith no claims yet

Pith reviewed 2026-05-13 16:54 UTC · model grok-4.3

classification 💻 cs.SE

keywords code reviewsemantic similaritypatch linkinggerritchrome extensionrelated changessoftware maintenance

0 comments

The pith

SmartPatchLinker ranks semantically related patches inside the Gerrit review interface using a local model.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

SmartPatchLinker is designed to find patches that are semantically related to the one under review, such as alternative solutions or overlapping modifications. These links often surface only days later, causing duplicated effort. The tool runs as a Chrome extension with a local backend that computes similarity scores on the fly when a reviewer opens a patch in Gerrit. Reviewers can adjust the search scope and see ranked suggestions with confidence scores without switching tools or installing server software. The authors evaluate both the tool's usefulness in practice and its usability for reviewers.

Core claim

SmartPatchLinker is implemented as a lightweight Chrome extension with a local inference backend and integrates with Gerrit to retrieve and rank semantically linked changes when a reviewer opens a patch. The tool allows reviewers to configure the search scope, view ranked candidates with confidence indicators, and examine related work without leaving their workflow or relying on server-side installations.

What carries the argument

A local inference backend that applies a semantic similarity model to rank candidate patches retrieved from Gerrit.

If this is right

Reviewers discover related patches earlier in the process, reducing duplicated effort.
Changes can be reviewed together without days of delay.
The open-source release allows teams to deploy the extension in their own Gerrit setups.
Evaluations show how the tool performs in usefulness and usability tests.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar local-matching extensions could be built for other code review platforms beyond Gerrit.
Accuracy of the semantic model is key; improvements in embedding quality would directly raise the value of the ranked list.
The approach separates client-side inference from server changes, which may ease adoption in security-conscious environments.

Load-bearing premise

The semantic similarity model used in the local inference backend produces rankings accurate enough to be useful to reviewers in real workflows.

What would settle it

A usability study that measures the fraction of suggested links that reviewers actually inspect and find relevant, compared against a baseline of no suggestions.

Figures

Figures reproduced from arXiv: 2604.04045 by Dong Wang, Islem Khemissi, Moataz Chouchen, Raula Gaikovina Kula.

**Figure 2.** Figure 2: SmartPatchLinker UI overview. and enables the tool to operate immediately on any public or private Gerrit instance after installation. The Chrome extension popup acts as the command center of the tool ( A ), allowing reviewers to configure the analysis scope by adjusting the lookback window (e.g., 2, 7, 14, or 30 days) and the number of results to retrieve (TopK), and to explicitly trigger the prediction … view at source ↗

**Figure 1.** Figure 1: shows an overview of the SmartPatchLinker. SmartPatchLinker Backend SmartPatchLinker UI SmartPatchLinker Model Gerrit Rest API Post request Top-k similar patches Fetch data Predict Return similarity scores [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 3.** Figure 3: Recall@K scores for SmartPatchLinker compared to Wang et al. [17] baselines. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 4.** Figure 4: Configuration: Time Window and Top-K [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: Top-K Results [5] Deepika Badampudi, Michael Unterkalmsteiner, and Ricardo Britto. 2023. Modern code reviews—survey of literature and practice. ACM Transactions on Software Engineering and Methodology 32, 4 (2023), 1–61. [6] Umut Cihan, Arda İçöz, Vahid Haratian, and Eray Tüzün. 2025. Evaluating Large Language Models for Code Review. arXiv preprint arXiv:2505.20206 (2025). [7] Siyue Feng, Wenqi Suo, Yuemin… view at source ↗

read the original abstract

In large software ecosystems, semantically related code changes, such as alternative solutions or overlapping modifications are often discovered only days after submission, leading to duplicated effort and delayed reviews. We present SmartPatchLinker, a browser based tool that supports the discovery of related patches directly within the code review interface. SmartPatchLinker is implemented as a lightweight Chrome extension with a local inference backend and integrates with Gerrit to retrieve and rank semantically linked changes when a reviewer opens a patch. The tool allows reviewers to configure the search scope, view ranked candidates with confidence indicators, and examine related work without leaving their workflow or relying on server-side installations. We perform both usefulness and usability evaluations to study how SmartPatchLinker can support reviewers during code review. SmartPatchLinker is open source, and its source code, Docker containers, and the replication package used in our evaluation are publicly available on GitHub at https://github.com/islem-kms/gerrit-chrome-extension . A video demonstrating the tool is also available online at https://drive.google.com/drive/folders/1MCcTj5OSlT7lHVBFMq5m9iatas2joaGb

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

SmartPatchLinker is a working open-source Chrome extension for ranking related patches inside Gerrit, but the paper gives no numbers on ranking quality or real-world impact.

read the letter

The paper describes SmartPatchLinker, a Chrome extension that ranks semantically related patches inside the Gerrit review interface using a local inference backend. The key point for you is that this is mainly an engineering contribution: a working, open-source tool that avoids server-side installs and lets reviewers configure the search on the fly. The implementation looks solid. It retrieves patches, computes similarity scores, displays ranked candidates with confidence indicators, and keeps everything lightweight. The authors made the source code, Docker containers, and replication package available on GitHub, which is a plus for anyone who wants to test or adapt it. They also include a demo video. The main weakness is the evaluation. Usefulness and usability studies are mentioned, but the text gives no concrete numbers on how well the rankings perform—no precision@K, no false positive rates, no baseline like simple file-name matching, and no details on the study design or participant feedback. Without that data, it's difficult to tell whether the tool actually reduces duplicated effort or just surfaces plausible but irrelevant links. This work is aimed at software engineering researchers and tool builders focused on code review processes. Someone working on similar integrations would get value from the architecture and the public artifacts, even if the empirical support needs strengthening. I would send this to peer review. The tool is concrete and the open-source release is genuine, so referees can provide targeted feedback on the missing metrics and model choices.

Referee Report

2 major / 1 minor

Summary. The paper presents SmartPatchLinker, a lightweight Chrome extension with a local inference backend that integrates with Gerrit to retrieve and rank semantically linked code changes when a reviewer opens a patch. Reviewers can configure search scope, view ranked candidates with confidence indicators, and examine related patches without leaving the workflow or requiring server-side setup. The authors state that usefulness and usability evaluations were performed, and provide a public GitHub repository with source code, Docker containers, and replication package.

Significance. If the semantic similarity rankings prove accurate in practice, the tool could reduce duplicated effort and delayed reviews in large software projects by embedding related-patch discovery directly in the code review interface. The open-source release and public replication package are positive factors that support reproducibility and potential adoption.

major comments (2)

[Abstract and Evaluation] Abstract and Evaluation section: The manuscript states that usefulness and usability evaluations were performed, yet provides no methods, metrics (e.g., precision@K, NDCG, inter-rater agreement), quantitative results, or baseline comparisons against real Gerrit patches. This directly undermines assessment of the central claim that the local inference backend produces rankings accurate enough to be useful to reviewers.
[Implementation] Implementation section: No details are given on the semantic similarity model (architecture, training data, or inference procedure), how confidence scores are derived, or any validation of ranking quality. Without these, it is impossible to judge whether the tool's core mechanism meets the accuracy threshold required for workflow integration.

minor comments (1)

[Abstract] The video demonstration link points to a Google Drive folder rather than a direct playable file; a direct link would improve accessibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We agree that additional details are needed to strengthen the manuscript and will incorporate the suggested expansions in the revised version.

read point-by-point responses

Referee: [Abstract and Evaluation] Abstract and Evaluation section: The manuscript states that usefulness and usability evaluations were performed, yet provides no methods, metrics (e.g., precision@K, NDCG, inter-rater agreement), quantitative results, or baseline comparisons against real Gerrit patches. This directly undermines assessment of the central claim that the local inference backend produces rankings accurate enough to be useful to reviewers.

Authors: We agree that the Evaluation section currently lacks the necessary methodological details and quantitative results. In the revised manuscript, we will expand this section to describe the evaluation methods in full, report the specific metrics used (including precision@K, NDCG, and inter-rater agreement where applicable), present the quantitative results obtained, and include baseline comparisons performed on real Gerrit patches. This will directly address the concern regarding the accuracy and usefulness of the rankings. revision: yes
Referee: [Implementation] Implementation section: No details are given on the semantic similarity model (architecture, training data, or inference procedure), how confidence scores are derived, or any validation of ranking quality. Without these, it is impossible to judge whether the tool's core mechanism meets the accuracy threshold required for workflow integration.

Authors: We acknowledge that the Implementation section does not currently provide sufficient technical details on the semantic similarity model. We will revise the manuscript to include a complete description of the model architecture, the training data employed, the inference procedure, the derivation of confidence scores, and the validation steps taken to assess ranking quality. These additions will allow readers to evaluate the suitability of the core mechanism for integration into the code review workflow. revision: yes

Circularity Check

0 steps flagged

No significant circularity: tool implementation with no derivation chain

full rationale

The paper presents SmartPatchLinker as a Chrome extension with a local inference backend for ranking semantically related patches in Gerrit reviews. No equations, parameter fittings, uniqueness theorems, or ansatzes are described that could reduce to self-defined inputs or self-citations by construction. Usefulness and usability evaluations are mentioned but treated as separate empirical steps rather than derived predictions. The work is self-contained as an engineering artifact whose claims rest on implementation details and external replication package availability, not on any load-bearing reduction to prior author results.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a software tool paper with no free parameters, axioms, or invented entities in a theoretical sense.

pith-pipeline@v0.9.0 · 5522 in / 875 out tokens · 29499 ms · 2026-05-13T16:54:52.936616+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages · 1 internal anchor

[1]

Muhammad Ahasanuzzaman, Muhammad Asaduzzaman, Chanchal K Roy, and Kevin A Schneider. 2016. Mining duplicate questions in stack overflow. InPro- ceedings of the 13th International Conference on Mining Software Repositories. 402–412

work page 2016
[2]

Ali Arabat and Mohammed Sayagh. 2024. An empirical study on cross-component dependent changes: A case study on the components of OpenStack.Empirical Software Engineering29, 5 (2024), 109

work page 2024
[3]

Ali Arabat, Mohammed Sayagh, and Jameleddine Hassine. 2025. An ML- based Approach to Predicting Software Change Dependencies.arXiv preprint arXiv:2508.05034(2025)

work page arXiv 2025
[4]

Alberto Bacchelli and Christian Bird. 2013. Expectations, outcomes, and chal- lenges of modern code review. In2013 35th International Conference on Software Engineering (ICSE). IEEE, 712–721. Figure 4: Configuration: Time Window and Top-K Figure 5: Top-K Results

work page 2013
[5]

Deepika Badampudi, Michael Unterkalmsteiner, and Ricardo Britto. 2023. Modern code reviews—survey of literature and practice.ACM Transactions on Software Engineering and Methodology32, 4 (2023), 1–61

work page 2023
[6]

Umut Cihan, Arda İçöz, Vahid Haratian, and Eray Tüzün. 2025. Evaluating Large Language Models for Code Review.arXiv preprint arXiv:2505.20206(2025)

work page arXiv 2025
[7]

Siyue Feng, Wenqi Suo, Yueming Wu, Deqing Zou, Yang Liu, and Hai Jin. 2024. Machine learning is all you need: A simple token-based approach for effective code clone detection. InProceedings of the IEEE/ACM 46th International Conference on Software Engineering. 1–13

work page 2024
[8]

Toshiki Hirao, Shane McIntosh, Akinori Ihara, and Kenichi Matsumoto. 2019. The review linkage graph for code review analytics: A recovery approach and empirical study. InProceedings of the 2019 27th ACM joint meeting on European software engineering conference and symposium on the foundations of software engineering. 578–589. Conference acronym ’XX, June ...

work page 2019
[9]

Junyi Lu, Lei Yu, Xiaojia Li, Li Yang, and Chun Zuo. 2023. Llama-reviewer: Ad- vancing code review automation with large language models through parameter- efficient fine-tuning. In2023 IEEE 34th International Symposium on Software Reliability Engineering (ISSRE). IEEE, 647–658

work page 2023
[10]

Anh Tuan Nguyen, Tung Thanh Nguyen, Tien N Nguyen, David Lo, and Cheng- nian Sun. 2012. Duplicate bug report detection with a combination of information retrieval and topic modeling. InProceedings of the 27th IEEE/ACM International Conference on Automated Software Engineering. 70–79

work page 2012
[11]

Chanathip Pornprasit and Chakkrit Tantithamthavorn. 2024. Fine-tuning and prompt engineering for large language models-based code review automation. Information and Software Technology175 (2024), 107523

work page 2024
[12]

Nikitha Rao, Bogdan Vasilescu, and Reid Holmes. 2025. From overload to insight: Bridging code search and code review with llms. InProceedings of the 33rd ACM International Conference on the Foundations of Software Engineering. 656–660

work page 2025
[13]

Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks.arXiv preprint arXiv:1908.10084(2019)

work page internal anchor Pith review Pith/arXiv arXiv 2019
[14]

Xunzhu Tang, Kisub Kim, Saad Ezzini, Yewei Song, Haoye Tian, Jacques Klein, and Tegawende Bissyande. 2025. Just-in-time detection of silent security patches. ACM Transactions on Software Engineering and Methodology(2025)

work page 2025
[15]

Patanamon Thongtanunam, Chakkrit Tantithamthavorn, Raula Gaikovina Kula, Norihiro Yoshida, Hajimu Iida, and Ken-ichi Matsumoto. 2015. Who should review my code? a file location-based code-reviewer recommendation approach for modern code review. In2015 IEEE 22nd international conference on software analysis, evolution, and reengineering (SANER). IEEE, 141–150

work page 2015
[16]

Panya Trakoolgerntong, Tao Xiao, Masanari Kondo, Chaiyong Ragkhitwetsagul, Morakot Choetkiertikul, Pattaraporn Sangaroonsilp, and Yasutaka Kamei. 2025. AILINKPREVIEWER: Enhancing Code Reviews with LLM-Powered Link Pre- views.arXiv preprint arXiv:2511.09223(2025)

work page arXiv 2025
[17]

Dong Wang, Raula Gaikovina Kula, Takashi Ishio, and Kenichi Matsumoto. 2021. Automatic patch linkage detection in code review using textual content and file location features.Information and Software Technology139 (2021), 106637

work page 2021
[18]

Qi Wang, Jindong Li, Shiqi Wang, Qianli Xing, Runliang Niu, He Kong, Rui Li, Guodong Long, Yi Chang, and Chengqi Zhang. 2024. Towards next- generation llm-based recommender systems: A survey and beyond.arXiv preprint arXiv:2410.19744(2024)

work page arXiv 2024
[19]

Yanming Yang, Ying Zou, Xing Hu, David Lo, Chao Ni, John Grundy, and Xin Xia. 2023. C 3: Code Clone-Based Identification of Duplicated Components. In Proceedings of the 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering. 1832–1843

work page 2023