ASE 2026 Overview

This year we had a paper accepted for publication at the IEEE/ACM International Conference on Automated Software Engineering (ASE 2026), the premier research conference on automated software engineering. Unfortunately the co-authors and I were unable to travel to attend the conference in-person in Munich, Germany this year, but this post summarizes our contribution:

Repository-Aware Metamorphic Relation Generation for Augmented Reality Applications using Large Language Models

Dibyendu Brinto Bose, Jiawei Qin, Chris Brown

• 🔍 Problem: Augmented reality (AR) applications are difficult to test because their behavior depends on unpredictable aspects such as changing physical environments, sensor data, camera input, spatial tracking, and user movement. Traditional testing cannot define the “correct” output for every situation. This paper explores whether large language models (LLMs) can generate metamorphic relations---rules describing how a programs output should change when its input is transformed---to create more practical AR test oracles.

• 🧪 Study: We analyzed 142 open-source Unity-based mobile AR repositories and prompted an LLM to generate metamorphic relations. Then we used a multi-agent debate process to compare, merge, or discard conflicting and redundant candidate relations and conducted a manual oracle study to observe the generated test cases.

• 📊 Findings: The hierarchical repository context (consisting of repository metadata, class relationships, and method details) had the highest coverage and more specific input and output transformations than the other approaches. We also observed the debate process detected conflicts or redundancies in 90.1% of candidate method-relation sets.

• 💡 Implication: Our results suggest that LLMs can contribute useful metamorphic test oracles for AR software, but their performance depends heavily on the quality and organization of the context received. Further studies are needed to confirm how well these generated relations detect real faults in deployed AR applications.

Unfortunately, we will not be able to attend this conference in person, but we are happy to answer questions and discuss this work. A pre-print of the paper can be found here. Please feel free to reach out if you would like to learn more about our approach, findings, or potential applications!