# Claims: Self-Admitted GenAI Usage in Open-Source Software Paper `tse26-genai-usage` (18 pages, `evals/papers/tse26-genai-usage.pdf`): 6 broad statements, 11 narrow claims, 65 rejected candidates. Generated from `claims.json` by `cea_claims.py render`. To change the record, edit `claims.json` and render again. ## Broad statements ### B1: rq_answer, page 7 > For the 1,292 GenAI mentions we analyzed, developers mainly used GenAI tools for code generation, natural language translation, and code refactoring. - Section: 3.2.3 Purposes of GenAI Usage, Summary RQ1 - Narrow claims: C1, C2 ### B2: rq_answer, page 7 > Source code and documentation files were the dominant generation targets. - Section: 3.2.3 Purposes of GenAI Usage, Summary RQ1 - Narrow claims: C3, C4 - Note: The text and Table 5 order the content types differently. The sentence names source code and documentation files as dominant, while Table 5 (page 7) prints Commit messages under Project metadata with 1,003, Documentation files with 106, and the six source-file codes with 47, 45, 39, 21, 12, and 12. ### B3: rq_answer, page 7 > Acknowledgment of GenAI usage was a common purpose, sometimes combined with warnings about possible negative implications. - Section: 3.2.3 Purposes of GenAI Usage, Summary RQ1 - Narrow claims: C5 ### B4: rq_answer, page 12 > Our results revealed that for most of the repositories analyzed, there was no significant change in code churn after GenAI adoption. - Section: 5.2 Results, Summary RQ3 - Narrow claims: C9, C10 ### B5: rq_answer, page 12 > Among the GenAI tasks identified in RQ1, generation tasks show a stronger impact on code churn than other tasks. - Section: 5.2 Results, Summary RQ3 - Narrow claims: C11 ### B6: conclusion, page 16 > We also observed recurring follow-up actions, including bug fixes, refactorings, reversions, and deletions, triggered by earlier GenAI-generated content. - Section: 9 Conclusion - Narrow claims: C6, C7, C8 - Note: The boxed answer for RQ1 states this main result without a quantitative part in numbers or in words (page 7, "Our analysis revealed patterns of corrective actions following code generation."), so the order passes over it and the conclusion sentence is quoted here, whose "recurring" rests on the counts in Section 3.2.1 on page 6. ## Narrow claims Grouped under each main result they serve, the result the paper states in the most places first. Within a result the claims stand in page order, because the paper puts no order on them. ### B4 (rq_answer, stated in 7 places): Our results revealed that for most of the repositories analyzed, there was no significant change in code churn after GenAI adoption. - C9, page 11, 1 of 2 for this result - C10, page 12, 1 of 2 for this result ### B5 (rq_answer, stated in 4 places): Among the GenAI tasks identified in RQ1, generation tasks show a stronger impact on code churn than other tasks. - C11, page 12, the only claim for this result ### B1 (rq_answer, stated in 3 places): For the 1,292 GenAI mentions we analyzed, developers mainly used GenAI tools for code generation, natural language translation, and code refactoring. - C1, page 5, 1 of 2 for this result - C2, page 5, 1 of 2 for this result ### B3 (rq_answer, stated in 2 places): Acknowledgment of GenAI usage was a common purpose, sometimes combined with warnings about possible negative implications. - C5, page 7, the only claim for this result ### B6 (conclusion, stated in 2 places): We also observed recurring follow-up actions, including bug fixes, refactorings, reversions, and deletions, triggered by earlier GenAI-generated content. - C6, page 6, 1 of 3 for this result - C7, page 6, 1 of 3 for this result - C8, page 6, 1 of 3 for this result ### B2 (rq_answer, stated in 1 place): Source code and documentation files were the dominant generation targets. - C3, page 7, 1 of 2 for this result - C4, page 7, 1 of 2 for this result ### Every claim #### C1: page 5, 3.2.1 GenAI-Assisted Tasks > Unsurprisingly, excluding PR-related activities, generation tasks were most common, with code generation being particularly prominent (105 instances). - Serves: B1 - Selection reason: B1 says developers mainly used GenAI tools for code generation. If code generation were not the most frequent task apart from PR descriptions, B1 would name the wrong task first and would need substantial revision. - Note: Table 3 (page 5) gives the code Code under the category Generation as 105, the same number as the sentence. Table 3 also gives PR description as 1,009, which the sentence excludes as a PR-related activity. #### C2: page 5, 3.2.1 GenAI-Assisted Tasks > Translation followed with 50 instances, while optimization and maintenance tasks accounted for 34 and 26 instances, respectively. - Part: Translation followed with 50 instances, while optimization tasks accounted for 34 instances. - Split from S1, with R1 - Serves: B1 - Selection reason: B1 says developers mainly used GenAI tools for natural language translation and code refactoring, and these are the counts behind the second and third place of that ranking. If translation and optimization did not rank where the sentence puts them, B1 would name the wrong tasks and would need substantial revision. - Note: B1 names code refactoring, and the only sentence with a number gives optimization, the category that holds it. Table 3 (page 5) prints the parts of both totals and no total: under Translation it gives Text 49 and Code 1, and under Optimization it gives Code refactoring 29 and Code improvement 5. The arithmetic is left to a later step. The sentence pairs its items with its numbers by "respectively", so the first number is optimization's and the second is maintenance's, which is how C2 and R1 divide it. #### C6: page 6, 3.2.1 GenAI-Assisted Tasks > We identified 20 mentions exhibiting this pattern of human actions triggered by an earlier GenAI action. - Serves: B6 - Selection reason: B6 says the study observed recurring follow-up actions triggered by earlier GenAI-generated content, and this count is what "recurring" rests on. If no mentions had shown the pattern, B6 would fail. - Note: No table or figure reports this number. #### C7: page 6, 3.2.1 GenAI-Assisted Tasks > The most common follow-up activity was to fix bugs in AI-generated code (9). - Serves: B6 - Selection reason: B6 names bug fixes first among the recurring follow-up actions. If no follow-up action had been a bug fix, B6 would have to drop the item it names first and would need substantial revision. - Note: No table or figure reports this number. #### C8: page 6, 3.2.1 GenAI-Assisted Tasks > In other cases, changes were reverted (1), AI-generated comments were deleted (2), or the generated code was commented out (1). - Part: In other cases, changes were reverted (1), AI-generated comments were deleted (2). - Split from S2, with R2 - Serves: B6 - Selection reason: B6 names reversions and deletions among the recurring follow-up actions. If neither had occurred, B6 would have to drop two of the four items it names and would need substantial revision. - Note: No table or figure reports these numbers. Table 6 (page 7) counts the purposes Revert 1 and Removal of Copilot comment 2, which is a different coding dimension. #### C3: page 7, 3.2.2 Content Types > Developers frequently use GenAI tools to modify source files (176 mentions). - Serves: B2 - Selection reason: B2 says source code files were a dominant generation target. If source files were not the content type with the largest count outside the commit messages of the single PR-description repository, B2 would fail. - Note: Table 5 (page 7) prints only the parts of this number, the six codes under Source files with 47, 45, 39, 21, 12, and 12, and no total. The arithmetic is left to a later step. #### C4: page 7, 3.2.2 Content Types > However, other file types, such as documentation and configuration files, were also targeted (135 mentions). - Serves: B2 - Selection reason: B2 says documentation files were a dominant generation target, and this is the only sentence that gives a number for them. If documentation files had barely been targeted, B2 would fail. - Note: B2 names documentation files, and this sentence gives the number for the category that holds them. Table 5 (page 7) prints the parts of the sentence's 135 and no total: Documentation files 106, Configuration files 24, and Resource files 5. The arithmetic is left to a later step. #### C5: page 7, 3.2.3 Purposes of GenAI Usage > Documentation and acknowledgment of GenAI usage emerged as the most frequent purpose. - Serves: B3 - Selection reason: B3 says acknowledgment of GenAI usage was a common purpose. If this category were not the most frequent one, B3 would fail. - Note: No sentence in the results or the discussion gives a count for this purpose. Table 6 (page 7) prints the codes of the category Documentation and Acknowledgment as Acknowledgement of usage 1,236, Acknowledge that the bug fix is related to AI-generated code 13, and Removal of Copilot comment 2, and gives no category total. #### C9: page 11, 5.2 Results > The average file-based code churn decreased from 0.17 to 0.06 with a significant difference (p < 0.05) and a medium effect (|δ| = 0.42), the average line-based churn decreased from 0.68 to 0.50 with a significant difference (p < 0.05) and a negligible effect (|δ| = 0.09). - Serves: B4 - Selection reason: B4 says there was no significant change in code churn after GenAI adoption for most repositories, and the discussion names the decreasing average churn as one of the two results it rests on (page 14, "The overall trend we observed pointed in the opposite direction, i.e., we noticed a decreasing average code churn."). If the averages had risen instead, B4 would fail. - Note: No table or figure reports these averages; Table 8 (page 11) reports the per-repository Wilcoxon effect sizes instead. #### C10: page 12, 5.2 Results > We observed that only 26 (file-based) and 30 (line-based) repositories showed significant code churn trends (p < 0.05). - Serves: B4 - Selection reason: B4 says most of the repositories analyzed showed no significant change in code churn, and these counts are what "most" rests on. If most repositories had shown a significant trend, B4 would fail. - Note: Table 9 (page 11) gives the same numbers: for file-based churn Sum sig. 26 and Not sig. 123, and for line-based churn Sum sig. 30 and Not sig. 119. #### C11: page 12, 5.2 Results > Across all models, significant discontinuities are concentrated in the generation category, including the generation of code, test data, test methods, regular expressions, and comments. In contrast, optimization and maintenance tasks do not show consistent patterns. - Serves: B5 - Selection reason: B5 says generation tasks show a stronger impact on code churn than other tasks, and this is the comparison behind it. If the significant discontinuities were spread evenly across the task categories, B5 would fail. - Note: Table 10 (page 11) gives the counts the sentence describes without numbers: for file-based churn Generation 16 significant and 68 not significant, Optimization 2 and 12, Maintenance 1 and 2; for line-based churn Generation 15 and 69, Optimization 4 and 10, Maintenance 2 and 1. The two sides of one comparison are kept together, so the sentence pair is one claim. ## Rejected candidates ### R3: page 1, Abstract > Using this concept as a lens to study how GenAI tools are integrated into OSS projects, we analyze a curated sample of more than 200,000 GitHub repositories, identifying 1,292 such self-admissions across 156 repositories in commit messages, code comments, and project documentation. - Reason: Every number in this sentence describes the study: the size of the repository sample, the number of mentions identified, and the number of repositories they come from. ### R4: page 1, Abstract > Using a mixed methods approach, we derive a taxonomy of 32 tasks, 10 content types, and 11 purposes associated with GenAI usage based on 1,292 qualitatively coded mentions. - Reason: Every number in this sentence describes the study: the sizes of the code book and the number of coded mentions. The taxonomy itself is a qualitative finding. ### R5: page 1, Abstract > We then analyze 13 documents with policies and usage guidelines for GenAI tools and conduct a developer survey to uncover the ethical, legal, and practical concerns behind them. - Reason: Its only number, the 13 documents analyzed, describes the material of the study, and the concerns it reports are a qualitative finding. ### R6: page 1, Abstract > Finally, we examine the longitudinal impact of GenAI adoption on code churn in 151 repositories with self-admitted GenAI usage and find no general increase, contradicting popular narratives on the impact of GenAI on software development. - Repeats: B4 - Reason: Repeats the main result that B4 states, and its 151 is the number of repositories analyzed. ### R7: page 2, 1 Introduction, contribution list > We introduce self-admitted GenAI usage as an empirical lens for studying GenAI adoption in open-source software and curate a dataset of 1,292 self-admitted GenAI usages across 156 GitHub repositories. - Reason: Every number in this sentence describes the study: the size of the curated dataset and the number of repositories it covers. ### R8: page 2, 1 Introduction, contribution list > Using a mixed-methods approach, we derive a taxonomy of GenAI usage comprising 32 development tasks, 10 content types, and 11 purposes, grounded in a qualitative analysis of the identified usages. - Reason: Every number in this sentence is a size of the code book, which describes the study, and the taxonomy itself is a qualitative finding. ### R9: page 2, 1 Introduction, contribution list > We empirically analyze how open-source projects govern GenAI usage by examining 13 policies and usage guidelines, contextualized through a developer survey. - Reason: Its only number, the 13 policies examined, describes the material of the study, and the sentence otherwise only announces an analysis. ### R10: page 2, 1 Introduction, contribution list > We assess the longitudinal impact of GenAI adoption on software evolution using a repository-level analysis of code churn in 151 projects, showing that GenAI adoption does not lead to a general increase in churn and that effects are stronger for generation tasks. - Repeats: B4, B5 - Reason: Repeats the two main results that B4 and B5 state, and its 151 is the number of projects analyzed. ### R16: page 2, 2.1 Repository Sampling > Our initial sample of GitHub projects contained 207,062 repositories distributed across Python (77,542), JavaScript (48,500), TypeScript (37,424), Java (25,160), and C# (18,436). - Reason: Gives the size of the initial repository sample, split by language, which describes the study. - Note: Figure 1 (page 3) reports the same size as 207,062 repositories in step 1 of the data collection process. ### R17: page 3, 2.1 Repository Sampling > Our final sample of GitHub repositories, obtained in February 2024, contained 14,785 GitHub repositories distributed across Java (5,060), C# (3,544), TypeScript (2,464), Python (1,875), and JavaScript (1,842). - Reason: Gives the size of the repository sample the results rest on, split by language, which describes the study. - Note: Table 1 (page 3) gives the same size in its caption as n = 14,785, and Figure 1 (page 3) reports 14,785 repositories in step 2. ### R18: page 4, 2.2 Identifying Self-Admitted GenAI Usages > Running the above retrieval process on all repositories yielded 3,004 mentions of GenAI tools: 1,572 in commit messages, 397 in source code comments, and 1,035 in documentation files. - Reason: Gives the number of mentions retrieved, split by artifact kind, which describes the material of the study. - Note: Figure 1 (page 3) reports the same total as 3,004 mentions in step 3 and gives no split by artifact kind. ### R19: page 4, 2.2 Identifying Self-Admitted GenAI Usages > We calculated a sample size of 341 mentions (of 3,004) to achieve estimates with a 95% confidence level and a 5% confidence interval. - Reason: Gives the size of the sample that the two authors labeled, which describes the study. ### R20: page 4, 2.2 Identifying Self-Admitted GenAI Usages > The inspection resulted in disagreement between the two authors for only 14 cases (4% of the sample). - Reason: Counts the disagreements in the coding procedure, which describes the study rather than a result. ### R21: page 4, 2.2 Identifying Self-Admitted GenAI Usages > The resulting 95% confidence interval was estimated to be (0.87, 0.95), indicating an “almost perfect” agreement. - Reason: Reports an inter-rater agreement score that supports the coding procedure, which describes the study. ### R22: page 4, 2.2 Identifying Self-Admitted GenAI Usages > In total, we identified 1,292 true-positive mentions of GenAI tools that were aligned with our inclusion criteria. We found true-positive mentions in 156 repositories (11 Python, 12 JavaScript, 37 TypeScript, 47 C#, and 49 Java repositories). - Reason: Gives the size of the coded material and the number of repositories it comes from, split by language, which describes the study. - Note: The two sentences are quoted together because the split by language belongs to the size that the first sentence states. Figure 1 (page 3) reports the same total as 1,292 usages in step 4. ### R23: page 4, 3.1 Method > A preliminary analysis identified 1,008 mentions related to Copilot for Pull Requests. - Reason: Describes the composition of the coded material and which mentions were deferred during coding, not a result. - Note: The quote leaves out the footnote marker attached to the last word. ### R24: page 4, 3.1 Method > To build the code book, two authors independently analyzed 284 GenAI mentions. - Reason: Gives the size of the material used to build the code book, which describes the study. - Note: The caption of Table 3 (page 5) gives the same number in n = 284 + 1,000 + 8 = 1,292. ### R1: page 5, 3.2.1 GenAI-Assisted Tasks > Translation followed with 50 instances, while optimization and maintenance tasks accounted for 34 and 26 instances, respectively. - Rejected part: Maintenance tasks accounted for 26 instances. (split from S1) - Reason: B1 would still stand, because it names code generation, natural language translation, and code refactoring, and maintenance is not one of the categories it names. - Note: Table 3 (page 5) prints the parts of the sentence's 26 and no total: Label revision 8, README revision 7, Document revision 4, Changelog revision 2, Prompt refinement 1, Color suggestion 1, Dependency upgrade 1, Version update 1, and Comment revision 1. The arithmetic is left to a later step. The sentence pairs its items with its numbers by "respectively", so the first number is optimization's and the second is maintenance's, which is how C2 and R1 divide it. ### R25: page 5, 3.1 Method > The analysis yielded “substantial” to “almost perfect” agreement levels on task (k = 0.81 − 0.89), content (k = 0.95 − 0.99), and purpose (k = 0.79 − 0.92), according to standard guidelines for interpreting k [20]. - Reason: Reports inter-rater agreement scores that support the coding procedure, which describes the study. ### R26: page 5, 3.2.1 GenAI-Assisted Tasks > Our analysis identified 32 distinct task categories in which developers use GenAI tools in their workflows. - Reason: Gives the number of codes in the task code book, which describes the study. ### R2: page 6, 3.2.1 GenAI-Assisted Tasks > In other cases, changes were reverted (1), AI-generated comments were deleted (2), or the generated code was commented out (1). - Rejected part: In other cases, the generated code was commented out (1). (split from S2) - Reason: B6 would still stand, because it names bug fixes, refactorings, reversions, and deletions, and commenting code out is not one of the actions it names. ### R27: page 6-7, 3.2.2 Content Types > Our analysis identified three main categories of AI-generated content in open-source software projects, organizing ten dis[...]tinct codes (see Table 5). - Reason: Gives the number of categories and codes in the content code book, which describes the study. - Note: The page break falls inside the word "distinct" and Tables 5 and 6 are printed between its two halves in text.txt. ### R30: page 6, 3.2.1 GenAI-Assisted Tasks > Of the 1,009 instances of code PR description in Table 3, 1,000 originated from a single repository and are discussed separately above. - Reason: Describes how the coded material is distributed over repositories, which describes the study. - Note: Table 3 (page 5) gives PR description as 1,009, the same number as the sentence. ### R31: page 6, 3.2.1 GenAI-Assisted Tasks > Among them, 13 referred to code that was initially generated using GenAI tools and then changed. - Reason: B6 would still stand, because this is the count for a subgroup of the 20 follow-up actions that C6 gives. ### R32: page 6, 3.2.1 GenAI-Assisted Tasks > In addition to the 13 human actions that followed AI code generation that we discussed above, we found seven human actions following the generation of configuration and validation files or an unclear role of the GenAI tool. - Reason: B6 would still stand, because this is the count for the remaining subgroup of the 20 follow-up actions that C6 gives. ### R11: page 7, 3.2.3 Purposes of GenAI Usage, Summary RQ1 > Our analysis revealed patterns of corrective actions following code generation. - Repeats: B6 - Reason: Repeats the main result that B6 states, without a quantitative part of its own. ### R12: page 7, 3.2.3 Purposes of GenAI Usage, Summary RQ1 > Another important purpose was regulation (see RQ2). - Breaks down: B3 - Reason: Adds a second purpose category to the purposes finding that B3 states, so it divides that result rather than stating one of its own. - Note: Table 6 (page 7) gives the codes of the category Guidance and Best Practices as Set example 25, Exclusion of usage within the project 18, and Regulation of usage within the project 10; the sentence gives no number. ### R28: page 7, 3.2.3 Purposes of GenAI Usage > Our analysis identified 11 different purposes for GenAI mentions in software projects, grouped into four main categories (see Table 6). - Reason: Gives the number of codes and categories in the purpose code book, which describes the study. ### R29: page 7, 3.2.3 Purposes of GenAI Usage > This manifested itself in several ways, such as offering guidance (53 mentions), flagging areas needing attention (23 mentions), and addressing GenAI limitations (4 mentions). - Reason: B3 would still stand, because it states that acknowledgment was a common purpose and these are the counts of the three other purpose categories. - Note: Table 6 (page 7) prints only the per-code counts and no category totals: for Guidance and Best Practices 25, 18, and 10; for Quality Assurance 11, 10, and 2; for GenAI Limitations 3 and 1. The arithmetic is left to a later step. ### R33: page 8, 4.1 Method > Using our sample of GenAI mentions, we found 28 mentions related to policies and usage guidelines around GenAI tool usage. - Reason: Gives the number of mentions that the RQ2 analysis rests on, which describes the study. ### R34: page 8, 4.1 Method > Table 7 presents detailed examples drawn from 13 documentation files and commit messages in 12 GitHub repositories, where the last column indicates the number of mentions identified in the software artifact. - Reason: Gives the number of policy documents and repositories that the RQ2 analysis rests on, which describes the study. - Note: Table 7 (page 9) lists thirteen policies, P1 to P13, and their last column gives the per-artifact mention counts 1, 1, 1, 9, 1, 1, 1, 1, 4, 1, 1, 1, and 5. ### R35: page 8, 4.1 Method > We were able to determine the email addresses of 30 contributors. In total, we received eight survey responses, which we analyzed using a combination of open coding and card sorting. - Reason: Gives the number of contributors contacted and the number of survey responses analyzed, which describes the study. - Note: The two sentences are quoted together because the response count is read against the number contacted. ### R36: page 8, 4.2.2 Developer Survey on GenAI Governance, General GenAI Usage Guidance > Five developers highlighted the necessity of regulating the usage of GenAI tools in software projects. - Reason: No broad statement states the main result that it would serve: the survey findings of RQ2 are qualitative and the record has no broad statement for them. - Note: The discussion builds on this count without a number (page 13, "Moreover, a majority of our survey participants agreed that the regulation of GenAI usage is necessary in open-source projects."), and the boxed answer for RQ2 (page 10) states only the kinds of policy found. ### R13: page 10, 4.2.2 Developer Survey on GenAI Governance, Summary RQ2 > We found 13 policies and guidelines on GenAI usage in open-source software projects, including strict policies prohibiting GenAI usage, policies requiring attribution, but also guidelines encouraging contributors to use GenAI, for example, for translating natural language text. - Reason: Its only number, the 13 policies analyzed, describes the material of the study, and the kinds of policy it lists are a qualitative finding. ### R38: page 10, 5.1 Method > We included 151 repositories with true positive GenAI mentions that did not prohibit the use of GenAI tools. - Reason: Gives the number of repositories that the RQ3 analysis rests on, which describes the study. - Note: The caption of Table 8 (page 11) gives the same size as n = 151, and Figure 1 (page 3) reports 151 repositories for the RDD analysis in step 4. ### R39: page 11, 5.1 Method > For the 149 repositories included, we identified four patterns, that we describe in the following: - Reason: Gives the number of repositories that the RDD analysis rests on, which describes the study. - Note: The caption of Table 9 (page 11) gives the same size as n = 149. ### R40: page 11, 5.2 Results > Of the 151 repositories with self-admitted GenAI usage, 119 had a significant difference in file-based churn, and 120 had a significant difference in line-based churn (p < 0.05). - Reason: B4 would still stand, because its first sentence summarises the RDD trend analysis, whose counts C10 gives from Table 9, while this sentence reports the Wilcoxon signed-rank comparison of Table 8. - Note: Table 8 (page 11) gives the same numbers in its Sum sig. column, 119 for file-based and 120 for line-based churn, against 32 and 31 not significant. The two analyses count significant repositories differently: Table 8 (page 11) gives 119 of 151 and 120 of 151 as significant under the Wilcoxon signed-rank test, and Table 9 (page 11) gives 26 of 149 and 30 of 149 as significant under the RDD analysis. The sentences following B4's first one, in the Summary RQ3 box on page 12, give 15 repositories with an upward trend and 15 with a downward trend, the numbers the RDD paragraph on page 12 reports. ### R41: page 11, 5.2 Results > Eleven repositories had an increasing file-based churn with a medium-to-large effect size, and 15 had an increasing line-based churn with a medium-to-large effect size. - Reason: B4 would still stand, because these are the counts for one effect-size subgroup of the repositories in Table 8. - Note: Table 8 (page 11) prints the parts of both numbers and no combined medium-to-large row: for file-based churn the increasing counts are medium 9 and large 2, and for line-based churn medium 7 and large 8. The arithmetic is left to a later step. ### R42: page 11, 5.2 Results > A decreasing churn was more common: 49 repositories had a decreasing file-based churn with a medium-to-large effect size, and 49 repositories had a decreasing line-based churn with a medium-to-large effect size. - Reason: B4 would still stand, because these are the counts for one effect-size subgroup of the repositories in Table 8. - Note: Table 8 (page 11) prints the parts of both numbers and no combined medium-to-large row: for file-based churn the decreasing counts are medium 23 and large 26, and for line-based churn medium 24 and large 25. The arithmetic is left to a later step. ### R14: page 12, 5.2 Results, Summary RQ3 > We did find 15 repositories with an overall upward trend in line-based code churn after the first GenAI mention. However, for most of them, the slope was negative. - Breaks down: B4 - Reason: Names the exception to the main result that B4 states, so it divides that result rather than stating one of its own. - Note: The second sentence starts with "However" and refers to "them", so it is quoted together with the sentence before it. Table 9 (page 11) prints the parts of the sentence's 15 and no total: for line-based churn Upward with #Positive slope 5 (16.7%) and #Negative slope 10 (33.3%). The arithmetic is left to a later step. ### R15: page 12, 5.2 Results, Summary RQ3 > In addition, we also found 15 repositories with a downward trend. - Breaks down: B4 - Reason: Names a subgroup of the repositories behind the main result that B4 states, so it divides that result rather than stating one of its own. - Note: Table 9 (page 11) prints the parts of the sentence's 15 and no total: for line-based churn Downward with #Positive slope 3 (10.0%) and #Negative slope 12 (40.0%). The arithmetic is left to a later step. ### R43: page 12, 5.2 Results > For file-based churn, an overall upward trend with a negative slope after the cutoff date was most common (12 repositories). - Reason: B4 would still stand, because this is the count for one of the four RDD patterns rather than the result as a whole. - Note: Table 9 (page 11) gives the same number for file-based churn, Upward trend with #Negative slope 12 (46.2%). ### R44: page 12, 5.2 Results > For line-based churn, an overall downward trend with a negative slope was most common (12 repositories). However, there were almost as many repositories (10) with an overall downward trend, but a positive slope after the cutoff date. - Reason: B4 would still stand, because these are the counts for two of the four RDD patterns rather than the result as a whole. - Note: The text and Table 9 put the 10 repositories in different patterns. The sentence gives them as an overall downward trend with a positive slope after the cutoff date, while Table 9 (page 11) prints 10 (33.3%) for line-based churn under Upward trend with #Negative slope and prints 3 (10.0%) under Downward trend with #Positive slope. Table 9 gives the sentence's 12 as Downward trend with #Negative slope 12 (40.0%). The second sentence starts with "However" and is quoted with the sentence before it. ### R45: page 12, 5.2 Results > We found 15 repositories with a significant upward trend in file-based churn and 15 with a significant upward trend in line-based churn. However, most of them had a negative slope. - Reason: B4 would still stand, because these are the counts for the repositories that are the exception to it. - Note: Table 9 (page 11) prints the parts of both numbers and no trend totals: for file-based churn Upward with #Positive 3 (11.5%) and #Negative 12 (46.2%), and for line-based churn Upward with #Positive 5 (16.7%) and #Negative 10 (33.3%). The arithmetic is left to a later step. The second sentence refers to "them" and is quoted with the sentence before it. ### R46: page 12, 5.2 Results > In addition, 11 projects had a significant downward trend in file-based churn and 15 had a significant downward trend in line-based churn. - Reason: B4 would still stand, because these are the counts for one of the trend directions rather than the result as a whole. - Note: Table 9 (page 11) prints the parts of both numbers and no trend totals: for file-based churn Downward with #Positive 4 (15.4%) and #Negative 7 (26.9%), and for line-based churn Downward with #Positive 3 (10.0%) and #Negative 12 (40.0%). The arithmetic is left to a later step. ### R47: page 12, 5.2 Results > We cannot conclude that there is a general trend toward increasing code churn. - Repeats: B4 - Reason: Repeats the main result that B4 states, without a number of its own. ### R48: page 12, 5.2 Results > Therefore, with our data and methodology, we cannot confirm this claim. - Repeats: B4 - Reason: Repeats the main result that B4 states, as a verdict on the GitClear claim. ### R49: page 12, 5.2 Results > In summary, this task-level analysis reveals that GenAI-assisted generation tasks appear more rework-prone than other forms of assistance. - Repeats: B5 - Reason: Repeats the main result that B5 states, without a number of its own. ### R37: page 13, 6.2 RQ2: Existing Guidelines for GenAI Usage > Moreover, a majority of our survey participants agreed that the regulation of GenAI usage is necessary in open-source projects. - Repeats: R36 - Reason: Repeats the count that R36 gives, in words, and no broad statement states the main result that it would serve. ### R50: page 13, 6.1 RQ1: Reasons for Mentioning GenAI Tools > Our analysis revealed that developers primarily use GenAI tools for code generation, natural language translation, and code refactoring. - Repeats: B1 - Reason: Repeats the main result that B1 states, without the counts. ### R51: page 13, 6.1 RQ1: Reasons for Mentioning GenAI Tools > In addition to code generation, we found that internationalization and natural language translation are common use cases for LLMs in open-source software projects (49 instances, see Table 3). - Reason: B1 would still stand, because this gives the count for the natural-language part of the translation category whose count C2 gives. - Note: The text gives the translation count differently in two places: this sentence gives 49 instances (page 13) and the results sentence of C2 gives 50 instances (page 5). Table 3 (page 5) prints Translation as Text 49 and Code 1. ### R52: page 13-14, 6.3 RQ3: Impact of GenAI Usage on Code Churn > Contrary to claims [...] in the GitClear report, which was extensively discussed in the software development community [36, 37], we did not find an increasing code churn after GenAI adoption. - Repeats: B4 - Reason: Repeats the main result that B4 states, without a number of its own. - Note: A footnote is printed between the two halves of the sentence in text.txt at the page break. ### R53: page 14, 6.3 RQ3: Impact of GenAI Usage on Code Churn > They found that approximately 54% of the generated code lines were integrated and only 2.5% of the generated snippets were later modified. - Reason: Reports the results of cited work, which the paper sets beside its own churn result, not a result of this study. ### R54: page 14, 6.3 RQ3: Impact of GenAI Usage on Code Churn > However, our RDD analysis revealed that three repositories (3/26, 11.5%) had a significant upward file-based code churn trend with a positive slope (p < 0.05). - Reason: B4 would still stand, because this is the count for one of the four RDD patterns and the paper presents it as the exception to that result. - Note: Table 9 (page 11) gives the same number for file-based churn, Upward trend with #Positive slope 3 (11.5%). ### R55: page 14, 6.3 RQ3: Impact of GenAI Usage on Code Churn > To contextualize this finding, we note that a larger proportion of repositories (12/26, 46.2%) showed an upward trend with a negative slope, i.e., churn increased, but the increase decelerated over time. - Reason: B4 would still stand, because this is the count for one of the four RDD patterns rather than the result as a whole. - Note: Table 9 (page 11) gives the same number for file-based churn, Upward trend with #Negative slope 12 (46.2%). ### R56: page 14, 6.3 RQ3: Impact of GenAI Usage on Code Churn > In addition, 26.9% (7/26) showed downward trends. - Reason: B4 would still stand, because this is the count for one of the four RDD patterns rather than the result as a whole. - Note: The text and Table 9 give the share of downward trends differently. The sentence gives 26.9% (7/26) for downward trends, while Table 9 (page 11) prints file-based Downward as #Positive slope 4 (15.4%) and #Negative slope 7 (26.9%), so the sentence's figure matches the negative-slope row alone. ### R57: page 15, 7.3 GenAI Impact on Software Development > The 2025 follow-up report [52], based on an expanded dataset of 211 million lines through 2024, reported a rise in code churn from 4.5% in 2023 to 5.7% in 2024, a 39.9% drop in refactoring, and a 17.1% increase in copy-pasted code. - Reason: Reports the results of cited work, which the paper sets beside its own churn result, not a result of this study. ### R58: page 16, 9 Conclusion > In our mixed-methods study design, we first mined more than 200,000 GitHub repositories, isolating 1,292 true-positive GenAI mentions across 156 projects. - Reason: Every number in this sentence describes the study: the size of the repository sample, the number of mentions, and the number of projects. ### R59: page 16, 9 Conclusion > Qualitative open coding of these instances and subsequent card sorting yielded taxonomies of 32 assisted tasks, 10 content types, and 11 usage purposes. - Reason: Every number in this sentence is a size of the code book, which describes the study, and the taxonomies themselves are a qualitative finding. ### R60: page 16, 9 Conclusion > We complemented this content analysis with a survey of project contributors and a systematic review of 13 project-level policies and guidelines. - Reason: Its only number, the 13 policies reviewed, describes the material of the study. ### R61: page 16, 9 Conclusion > In addition, we performed a regression-discontinuity (RDD) analysis of code churn in the 149 repositories that contained sufficient data to study the impact of GenAI adoption on open source software projects. - Reason: Its only number, the 149 repositories, is the size of the material the RDD analysis rests on, which describes the study. ### R62: page 16, 9 Conclusion > The RQ1 results show that developers most often use GenAI for code generation, natural-language translation, and refactoring—with explicit acknowledgment as the dominant purpose. - Repeats: B1, B3 - Reason: Repeats the two main results that B1 and B3 state, without the counts. ### R63: page 16, 9 Conclusion > Together with our RQ3 task-level analysis, which shows that generation tasks are more rework-prone than optimization or maintenance, this suggests that GenAI output has a provisional nature. - Repeats: B5 - Reason: Repeats the main result that B5 states, as the ground for an implication. ### R64: page 16, 9 Conclusion > While acknowledgment of GenAI usage is common (RQ1), contextual metadata such as prompts, model versions, or scope of generation is rarely recorded. - Reason: Cannot tell what the frequency word "rarely" rests on, so it is recorded here for the checker. - Note: The paper gives no count for how often the generation context is documented. The discussion states the same in words (page 13, "A more standardized approach for documenting GenAI contributions is required, since most self-admitted GenAI usages did not document the generation context beyond brief summaries."), and the codes of Table 6 (page 7) count purposes, not whether a mention records prompts, model versions, or the scope of generation. ### R65: page 16, 9 Conclusion > The RQ3 results show no systematic increase in code churn after GenAI adoption, contradicting prominent industry claims. - Repeats: B4 - Reason: Repeats the main result that B4 states, without a number of its own.