arXiv subjects

Aaron Imani

Publications and source records attributed to Aaron Imani.

7 recordsLinked to original sources

arXiv2026

ProjAgent: Procedural Similarity Retrieval for Repository-Level Code Generation

Repository-level code generation requires implementing target functions while accounting for complex cross-file dependencies and project-specific conventions. Existing retrieval methods predominantly rely on lexical, structural, or semantic similarity, often overlooking repository functions that implement similar procedural logic despite differing in identifiers or application domains. We propose ProjAgent, a repository-level code generation system that introduces procedural similarity as an explicit retrieval signal. ProjAgent decomposes the target function into intermediate reasoning steps and employs an agentic workflow to retrieve repository functions that exhibit similar procedural behavior at each step. The retrieved procedural context is integrated with conventional semantic retrieval to construct a richer repository context for code generation. ProjAgent further incorporates a conservative static-analysis feedback loop that iteratively repairs generated code using compiler and static-analysis feedback. Evaluated on REPOCOD, ProjAgent achieves 41.14% Pass@1, outperforming existing retrieval-based baselines. These results demonstrate that procedural similarity is an effective and previously unexplored retrieval dimension for repository-level code generation.

cs.SE↗

arXiv2026

From Anatomy to Smells: An Empirical Study of SKILL.md in Agent Skills

Agent Skills provide on-demand domain knowledge to LLM agents without requiring model retraining. Each Agent Skill is defined by a mandatory SKILL$.$md file containing metadata and an unstructured Markdown body whose contents are left entirely to the skill author. Despite the rapid adoption of Agent Skills, little is known about how these files are authored or whether existing authoring guidelines are followed in practice. In this paper, we present the first systematic study of SKILL$.$md files as a software artifact. We qualitatively analyze 238 real-world skills and derive a taxonomy of 13 higher-level and 44 lower-level semantic components. We then conduct a multivocal literature review of 29 sources to identify best practices for authoring SKILL$.$md files and introduce skill smells as violations of these practices. Finally, we develop an automated detector and apply it to real-world skills, finding that over 99% of SKILL$.$md files contain at least one skill smell, and once introduced, skill smells rarely disappear as skills evolve. These findings reveal a substantial gap between recommended and actual authoring practices, motivating the development of automated techniques to remediate skill smells while increasing developer awareness of this emerging quality issue.

cs.SE↗

arXiv2025

Inside Out: Uncovering How Comment Internalization Steers LLMs for Better or Worse

While comments are non-functional elements of source code, Large Language Models (LLM) frequently rely on them to perform Software Engineering (SE) tasks. Yet, where in the model this reliance resides, and how it affects performance, remains poorly understood. We present the first concept-level interpretability study of LLMs in SE, analyzing three tasks - code completion, translation, and refinement - through the lens of internal comment representation. Using Concept Activation Vectors (CAV), we show that LLMs not only internalize comments as distinct latent concepts but also differentiate between subtypes such as Javadocs, inline, and multiline comments. By systematically activating and deactivating these concepts in the LLMs' embedding space, we observed significant, model-specific, and task-dependent shifts in performance ranging from -90% to +67%. Finally, we conducted a controlled experiment using the same set of code inputs, prompting LLMs to perform 10 distinct SE tasks while measuring the activation of the comment concept within their latent representations. We found that code summarization consistently triggered the strongest activation of comment concepts, whereas code completion elicited the weakest sensitivity. These results open a new direction for building SE tools and models that reason about and manipulate internal concept representations rather than relying solely on surface-level input.

cs.SE↗

arXiv2024

Prompting in the Wild: An Empirical Study of Prompt Evolution in Software Repositories

The adoption of Large Language Models (LLMs) is reshaping software development as developers integrate these LLMs into their applications. In such applications, prompts serve as the primary means of interacting with LLMs. Despite the widespread use of LLM-integrated applications, there is limited understanding of how developers manage and evolve prompts. This study presents the first empirical analysis of prompt evolution in LLM-integrated software development. We analyzed 1,262 prompt changes across 243 GitHub repositories to investigate the patterns and frequencies of prompt changes, their relationship with code changes, documentation practices, and their impact on system behavior. Our findings show that developers primarily evolve prompts through additions and modifications, with most changes occurring during feature development. We identified key challenges in prompt engineering: only 21.9% of prompt changes are documented in commit messages, changes can introduce logical inconsistencies, and misalignment often occurs between prompt changes and LLM responses. These insights emphasize the need for specialized testing frameworks, automated validation tools, and improved documentation practices to enhance the reliability of LLM-integrated applications.

cs.SE↗

arXiv2024

Investigating the Impact of Code Comment Inconsistency on Bug Introducing

Code comments are essential for clarifying code functionality, improving readability, and facilitating collaboration among developers. Despite their importance, comments often become outdated, leading to inconsistencies with the corresponding code. This can mislead developers and potentially introduce bugs. Our research investigates the impact of code-comment inconsistency on bug introduction using large language models, specifically GPT-3.5. We first compare the performance of the GPT-3.5 model with other state-of-the-art methods in detecting these inconsistencies, demonstrating the superiority of GPT-3.5 in this domain. Additionally, we analyze the temporal evolution of code-comment inconsistencies and their effect on bug proneness over various timeframes using GPT-3.5 and Odds ratio analysis. Our findings reveal that inconsistent changes are around 1.5 times more likely to lead to a bug-introducing commit than consistent changes, highlighting the necessity of maintaining consistent and up-to-date comments in software development. This study provides new insights into the relationship between code-comment inconsistency and software quality, offering a comprehensive analysis of its impact over time, demonstrating that the impact of code-comment inconsistency on bug introduction is highest immediately after the inconsistency is introduced and diminishes over time.

cs.SE↗

arXiv2024

Context Conquers Parameters: Outperforming Proprietary LLM in Commit Message Generation

Commit messages provide descriptions of the modifications made in a commit using natural language, making them crucial for software maintenance and evolution. Recent developments in Large Language Models (LLMs) have led to their use in generating high-quality commit messages, such as the Omniscient Message Generator (OMG). This method employs GPT-4 to produce state-of-the-art commit messages. However, the use of proprietary LLMs like GPT-4 in coding tasks raises privacy and sustainability concerns, which may hinder their industrial adoption. Considering that open-source LLMs have achieved competitive performance in developer tasks such as compiler validation, this study investigates whether they can be used to generate commit messages that are comparable with OMG. Our experiments show that an open-source LLM can generate commit messages that are comparable to those produced by OMG. In addition, through a series of contextual refinements, we propose lOcal MessagE GenerAtor (OMEGA) , a CMG approach that uses a 4-bit quantized 8B open-source LLM. OMEGA produces state-of-the-art commit messages, surpassing the performance of GPT-4 in practitioners' preference.

cs.SE↗

arXiv2024

Does Documentation Matter? An Empirical Study of Practitioners' Perspective on Open-Source Software Adoption

In recent years, open-source software (OSS) has become increasingly prevalent in developing software products. While OSS documentation is the primary source of information provided by the developers' community about a product, its role in the industry's adoption process has yet to be examined. We conducted semi-structured interviews and an online survey to provide insight into this area. Based on interviews and survey insights, we developed a topic model to collect relevant information from OSS documentation automatically. Additionally, according to our survey responses regarding challenges associated with OSS documentation, we propose a novel information augmentation approach, DocMentor, by combining OSS documentation corpus TF-IDF scores and ChatGPT. Through explaining technical terms and providing examples and references, our approach enhances the documentation context and improves practitioners' understanding. Our tool's effectiveness is assessed by surveying practitioners.

cs.SE↗