F. Khomh

RefineAct: Automatic Runtime Verification of LLM Agent Actions

CRAYON: Supporting the Elicitation of Pluralistic Normative Requirements for Autonomous Systems

Normative requirements (NRs) capturing social, legal, ethical, empathetic, and cultural (SLEEC) expectations are increasingly critical for autonomous systems operating in humancentered environments. Governance frameworks such as the EU AI Act articulate these expectations as high-level principles, but translating them...

The relationship between different python argument-passing mechanisms and fixes: An empirical study

Evidence is found that keyword-based argument passing is less defect-prone than positional argument passing, and this is not affected by size-related confounding factors.

Towards Optimizing SQL Generation via LLM Routing

This paper introduces the first LLM routing approach for Text-to-SQL, which dynamically selects the most cost-effective LLM capable of generating accurate SQL for each query.

Revisiting Feedback-Driven LLM Code Repair: A Replication and Exploratory Java Extension

Since the advent of Large Language Models (LLMs), practitioners have increasingly leveraged them to support their software engineering tasks, including automated code repair, showing promising results. Yet, concerns regarding reproducibility and generalizability remain largely unexplored. To further evaluate these conc...

An empirical study on method-level performance evolution in open-source Java projects

A large-scale empirical study analyzing performance evolution in 15 mature open-source Java projects hosted on GitHub, showing that different project domains and scales exhibit distinct performance characteristics, which suggest the need for domain-specific performance management strategies that account for the project...

AgentInspect: Diagnosing Behavioral Failures in Artificial Intelligence Agents

This work developed AgentInspect, a framework that automatically detects six types of behavioral failures in LangChain-based AI agents by analyzing their execution trajectories across three evaluation settings: a baseline setting using real tool responses, a simulated setting incorporating synthetic tool responses, and...

Securing Time Integrity in Energy IoT Against Clock Drift and Y2K38 Failures

STGAT (Spatio-Temporal Graph Attention Network), a clock-dynamics-aware anomaly detection solution that jointly models temporal distortion and inter-device consistency in energy IoT systems, is introduced.

Impact in Software Engineering Activities After One Year of COVID-19 Restrictions for Startups and Established Companies

Investigating the impacts of COVID-19 on software development activities after one year of the pandemic restrictions found that most respondents did not observe a significant impact, and software startups and established companies were affected differently.

LLM Serving in the Wild: An Empirical Study of Frameworks, Methods, and System Designs

This study identifies and analyze five LLM-specific frameworks and examines how these frameworks and techniques are adopted individually and in combination, how adoption varies across categories of LLMs, and how repositories differ in intent, focus, use case, and architectural design.

When Code Authors Are Agents: A Large-Scale Study of Human–Agent Collaboration in Pull Requests

Large-scale human–agent collaboration is investigated by comparing 40,214 pull requests across 2,807 GitHub repositories, including 33,596 agent-authored PRs from five autonomous coding agents and 6,618 human-authored PRs, revealing a socio-technical trade-off.

How Do AI Coding Agents Contribute to Software Development? an Empirical Study of Agentic Pull Requests

This study characterize agentic pull requests (PR) in comparison to human generated PRs and to examine how their properties change across different stages of the development lifecycle, offering a more nuanced understanding of their benefits and limitations in real-world practices.