LLM Code Generation + LLM-Based Agents for GitHub Issue Enhancement

A﻿ Training Program on the Development, Deployment and Servicing of Artificial Intelligence-based Software Systems

CREATE SE4AI

Trainee Talks is a monthly webinar series featuring CREATE SE4AI trainees sharing their research on software engineering and AI.

Join us on Wednesday, September 30th at 3:00 PM EDT when 2 CREATE PhD students will deliver their research talks. First, we will hear from Fazle Rabbi of Concordia University. Fazle will present his work on "How Robust Is LLM Code Generation Across Programming Languages?"Immediately following, Pouya Fathollazadeh from Queen's University will speak on the topic of "LLM-Based Agents for GitHub Issue Enhancement: A Comprehensive Empirical Study".Each trainee will speak for 20 minutes, followed by a 10 min Q&amp;A session.

How Robust Is LLM Code Generation Across Programming Languages?Abstract:Large Language Models (LLMs) have shown strong capabilities in automated code generation, yet their reliability can vary substantially across programming languages. This work examines the robustness of LLM-generated code from a multi-language perspective, investigating how model performance changes across languages with different characteristics and levels of representation in training data. The study evaluates the correctness and robustness of generated programs under varying conditions and identifies language-dependent differences in model behavior. The findings highlight some limitations in current LLM-based code generation systems and emphasize the need for broader, language-aware evaluation when assessing their reliability for practical software development.

LLM-Based Agents for GitHub Issue Enhancement: A Comprehensive Empirical StudyAbstract:GitHub issue reports are essential for software development, yet many are incomplete or unclear, making them difficult for maintainers to act on. This talk presents IssueTuner, an agentic approach that automatically improves low-quality GitHub issues through an iterative workflow. The study examines how reliably issue quality can be assessed and how agent architecture affects enhancement performance, and whether the approach can scale to large numbers of real-world issues.

LLM Code Generation + LLM-Based Agents for GitHub Issue Enhancement

More events from CREATE SE4AI

Discover more events from CREATE SE4AI, from Science & Tech to other experiences you might love.

LLM Code Generation + LLM-Based Agents for GitHub Issue Enhancement

More events from CREATE SE4AI

Discover more events from CREATE SE4AI, from Science & Tech to other experiences you might love.

You might also like...

Browse more events with different dates, prices, and formats to find your next great experience.