arXiv subjects

Mohamed Abdelrazek

Publications and source records attributed to Mohamed Abdelrazek.

At least 19 recordsLinked to original sources

arXiv2026

When Does An Extra View Help? Adapting Single-View 3D Reconstruction with Extra Imagery

Reconstruction of 3D objects from a single image is a challenging research problem in computer vision. The key challenge is the lack of critical information from viewpoints to complete 3D structures. Using an additional view may help to resolve the issue. However, there is no mechanism that can integrate the extra view into the single-view 3D reconstruction principle. We address this challenge by proposing ASV3D, a framework for adapting single-view 3D object reconstruction to test-time data with support from one additional image. We introduce two adaptation strategies: (i) a zero-shot adaptation scheme that leverages the auxiliary image to improve the reconstruction quality of an object without retraining, and (ii) an optimised adaptation scheme that further enhances visual fidelity and cross-view consistency via contrastive learning. We apply our ASV3D to improve two state-of-the-art single-view 3D reconstruction pipelines on both benchmark and real-world datasets. Results demonstrate that our approach consistently improves reconstruction accuracy and robustness under unconstrained multi-view inputs, outperforming the baselines in both quantitative metrics and human preference. We publish our code and the real-world object dataset in our project page at https://github.com/YNhuHuynh/ASV3D/tree/main.

cs.CV↗

arXiv2026

Seeing Before Generating: Object Perception Enhances Single-View 3D Reconstruction

The relationship between object perception and reconstruction is well established in human vision, yet remains underexplored in computer vision. In this paper, we demonstrate that learnt object perception can significantly enhance 3D reconstruction. Focusing on the challenging task of single-view 3D object reconstruction, we propose a method that leverages perceptual signals extracted from pretrained perception models capturing semantic and geometric information to drive the reconstruction of an object from its single image. Our approach is model-agnostic and can be integrated into various reconstruction methods in a plug-and-play manner. Experiments with two state-of-the-art single-view 3D reconstruction pipelines in a benchmark dataset show consistent and substantial improvements achieved by our method, validating the effectiveness of incorporating perception into generation. We provide in-depth analysis of various aspects of our method and its application. Our project page is at https://ynhuhuynh.github.io/perception-3d/.

cs.CV↗

arXiv2026

Hidden in Thought: Transferable Chain-of-Thought Artifacts Induce Harmful Behavior

We investigate whether harmful chain-of-thought (CoT) traces from compromised language models can transfer unsafe behaviour and be distilled into reusable jailbreak attacks. Using an emergent-misalignment organism and a refusal-ablated jailbroken organism, we transplant harmful CoTs into $29$ open-source and $5$ closed-source targets. Transferred traces raise harmful-response rates above $80\%$ on the most vulnerable open-source models, while semantically mismatched CoTs fail entirely. LLooM concept mining identifies four recurring components of harmful reasoning: proceduralisation, ethical decoupling, evasion, and target--vulnerability framing. Distilling these patterns into reusable system prompts produces effective black-box jailbreaks, outperforming direct CoT transplantation on strongly aligned models by up to an order of magnitude, including a $10\times$ improvement on GPT-4.1 AdvBench. Reasoning-enabled models are more than twice as vulnerable, and output-side safeguards such as Llama-Guard~3 frequently miss harmful generations. Our results show that harmful reasoning transfers at both the trace and pattern levels, motivating defences that evaluate reasoning context in addition to final outputs.

cs.CR↗

arXiv2025

Ensuring Robustness in ML-enabled Software Systems: A User Survey

Ensuring robustness in ML-enabled software systems requires addressing critical challenges, such as silent failures, out-of-distribution (OOD) data, and adversarial attacks. Traditional software engineering practices, which rely on predefined logic, are insufficient for ML components that depend on data and probabilistic decision-making. To address these challenges, we propose the ML-On-Rails protocol, a unified framework designed to enhance the robustness and trustworthiness of ML-enabled systems in production. This protocol integrates key safeguards such as OOD detection, adversarial attack detection, input validation, and explainability. It also includes a model-to-software communication framework using HTTP status codes to enhance transparency in reporting model outcomes and errors. To align our approach with real-world challenges, we conducted a practitioner survey, which revealed major robustness issues, gaps in current solutions, and highlighted how a standardised protocol such as ML-On-Rails can improve system robustness. Our findings highlight the need for more support and resources for engineers working with ML systems. Finally, we outline future directions for refining the proposed protocol, leveraging insights from the survey and real-world applications to continually enhance its effectiveness.

cs.SE↗

arXiv2024

Large language models for generating rules, yay or nay?

Engineering safety-critical systems such as medical devices and digital health intervention systems is complex, where long-term engagement with subject-matter experts (SMEs) is needed to capture the systems' expected behaviour. In this paper, we present a novel approach that leverages Large Language Models (LLMs), such as GPT-3.5 and GPT-4, as a potential world model to accelerate the engineering of software systems. This approach involves using LLMs to generate logic rules, which can then be reviewed and informed by SMEs before deployment. We evaluate our approach using a medical rule set, created from the pandemic intervention monitoring system in collaboration with medical professionals during COVID-19. Our experiments show that 1) LLMs have a world model that bootstraps implementation, 2) LLMs generated less number of rules compared to experts, and 3) LLMs do not have the capacity to generate thresholds for each rule. Our work shows how LLMs augment the requirements' elicitation process by providing access to a world model for domains.

cs.SE↗

arXiv2024

Quantifying Manifolds: Do the manifolds learned by Generative Adversarial Networks converge to the real data manifold

This paper presents our experiments to quantify the manifolds learned by ML models (in our experiment, we use a GAN model) as they train. We compare the manifolds learned at each epoch to the real manifolds representing the real data. To quantify a manifold, we study the intrinsic dimensions and topological features of the manifold learned by the ML model, how these metrics change as we continue to train the model, and whether these metrics convergence over the course of training to the metrics of the real data manifold.

cs.LG↗

arXiv2024

LLMs for Test Input Generation for Semantic Caches

Large language models (LLMs) enable state-of-the-art semantic capabilities to be added to software systems such as semantic search of unstructured documents and text generation. However, these models are computationally expensive. At scale, the cost of serving thousands of users increases massively affecting also user experience. To address this problem, semantic caches are used to check for answers to similar queries (that may have been phrased differently) without hitting the LLM service. Due to the nature of these semantic cache techniques that rely on query embeddings, there is a high chance of errors impacting user confidence in the system. Adopting semantic cache techniques usually requires testing the effectiveness of a semantic cache (accurate cache hits and misses) which requires a labelled test set of similar queries and responses which is often unavailable. In this paper, we present VaryGen, an approach for using LLMs for test input generation that produces similar questions from unstructured text documents. Our novel approach uses the reasoning capabilities of LLMs to 1) adapt queries to the domain, 2) synthesise subtle variations to queries, and 3) evaluate the synthesised test dataset. We evaluated our approach in the domain of a student question and answer system by qualitatively analysing 100 generated queries and result pairs, and conducting an empirical case study with an open source semantic cache. Our results show that query pairs satisfy human expectations of similarity and our generated data demonstrates failure cases of a semantic cache. Additionally, we also evaluate our approach on Qasper dataset. This work is an important first step into test input generation for semantic applications and presents considerations for practitioners when calibrating a semantic cache.

cs.SE↗

arXiv2024

ML-On-Rails: Safeguarding Machine Learning Models in Software Systems A Case Study

Machine learning (ML), especially with the emergence of large language models (LLMs), has significantly transformed various industries. However, the transition from ML model prototyping to production use within software systems presents several challenges. These challenges primarily revolve around ensuring safety, security, and transparency, subsequently influencing the overall robustness and trustworthiness of ML models. In this paper, we introduce ML-On-Rails, a protocol designed to safeguard ML models, establish a well-defined endpoint interface for different ML tasks, and clear communication between ML providers and ML consumers (software engineers). ML-On-Rails enhances the robustness of ML models via incorporating detection capabilities to identify unique challenges specific to production ML. We evaluated the ML-On-Rails protocol through a real-world case study of the MoveReminder application. Through this evaluation, we emphasize the importance of safeguarding ML models in production.

cs.SE↗

arXiv2024

Seven Failure Points When Engineering a Retrieval Augmented Generation System

Software engineers are increasingly adding semantic search capabilities to applications using a strategy known as Retrieval Augmented Generation (RAG). A RAG system involves finding documents that semantically match a query and then passing the documents to a large language model (LLM) such as ChatGPT to extract the right answer using an LLM. RAG systems aim to: a) reduce the problem of hallucinated responses from LLMs, b) link sources/references to generated responses, and c) remove the need for annotating documents with meta-data. However, RAG systems suffer from limitations inherent to information retrieval systems and from reliance on LLMs. In this paper, we present an experience report on the failure points of RAG systems from three case studies from separate domains: research, education, and biomedical. We share the lessons learned and present 7 failure points to consider when designing a RAG system. The two key takeaways arising from our work are: 1) validation of a RAG system is only feasible during operation, and 2) the robustness of a RAG system evolves rather than designed in at the start. We conclude with a list of potential research directions on RAG systems for the software engineering community.

cs.SE↗

arXiv2023

6DVF: Data Visualisation Framework for mHealth Apps

The widespread of data visualisation tools on smartphones has provided end users an easy way to track their health data, leading designers to put more effort into delivering suitable visualisations. Both academia and industry have developed several frameworks to guide the creation of informative and well-designed charts, such as the visualisation and design framework and Google Material Design. Despite the typical focus on design and chart types in these existing frameworks, our study highlights the need to incorporate additional components when developing data visualisations. The needs of non-expert users, the nature of the data being represented, and the mobile environment are often not prioritised in these frameworks, leading to visualisations that do not meet user needs and expectations. To address these issues, we propose our Six-Dimensions Data Visualisation Framework (6DVF) to assist in the design and evaluation of visualisations on mobile devices. Finally, we present our initial findings from a designer evaluation experiment.

cs.HC↗

arXiv2023

Advancing Requirements Engineering through Generative AI: Assessing the Role of LLMs

Requirements Engineering (RE) is a critical phase in software development including the elicitation, analysis, specification, and validation of software requirements. Despite the importance of RE, it remains a challenging process due to the complexities of communication, uncertainty in the early stages and inadequate automation support. In recent years, large-language models (LLMs) have shown significant promise in diverse domains, including natural language processing, code generation, and program understanding. This chapter explores the potential of LLMs in driving RE processes, aiming to improve the efficiency and accuracy of requirements-related tasks. We propose key directions and SWOT analysis for research and development in using LLMs for RE, focusing on the potential for requirements elicitation, analysis, specification, and validation. We further present the results from a preliminary evaluation, in this context.

cs.SE↗

arXiv2023

An empirical study of automatic wildlife detection using drone thermal imaging and object detection

Artificial intelligence has the potential to make valuable contributions to wildlife management through cost-effective methods for the collection and interpretation of wildlife data. Recent advances in remotely piloted aircraft systems (RPAS or ``drones'') and thermal imaging technology have created new approaches to collect wildlife data. These emerging technologies could provide promising alternatives to standard labourious field techniques as well as cover much larger areas. In this study, we conduct a comprehensive review and empirical study of drone-based wildlife detection. Specifically, we collect a realistic dataset of drone-derived wildlife thermal detections. Wildlife detections, including arboreal (for instance, koalas, phascolarctos cinereus) and ground dwelling species in our collected data are annotated via bounding boxes by experts. We then benchmark state-of-the-art object detection algorithms on our collected dataset. We use these experimental results to identify issues and discuss future directions in automatic animal monitoring using drones.

cs.CV↗

arXiv2023

Requirements Engineering Framework for Human-centered Artificial Intelligence Software Systems

[Context] Artificial intelligence (AI) components used in building software solutions have substantially increased in recent years. However, many of these solutions focus on technical aspects and ignore critical human-centered aspects. [Objective] Including human-centered aspects during requirements engineering (RE) when building AI-based software can help achieve more responsible, unbiased, and inclusive AI-based software solutions. [Method] In this paper, we present a new framework developed based on human-centered AI guidelines and a user survey to aid in collecting requirements for human-centered AI-based software. We provide a catalog to elicit these requirements and a conceptual model to present them visually. [Results] The framework is applied to a case study to elicit and model requirements for enhancing the quality of 360 degree~videos intended for virtual reality (VR) users. [Conclusion] We found that our proposed approach helped the project team fully understand the human-centered needs of the project to deliver. Furthermore, the framework helped to understand what requirements need to be captured at the initial stages against later stages in the engineering process of AI-based software.

cs.SE↗

arXiv2023

Requirements Elicitation and Modelling of Artificial Intelligence Systems: An Empirical Study

Artificial Intelligence (AI) systems have gained significant traction in the recent past, creating new challenges in requirements engineering (RE) when building AI software systems. RE for AI practices have not been studied much and have scarce empirical studies. Additionally, many AI software solutions tend to focus on the technical aspects and ignore human-centered values. In this paper, we report on a case study for eliciting and modeling requirements using our framework and a supporting tool for human-centred RE for AI systems. Our case study is a mobile health application for encouraging type-2 diabetic people to reduce their sedentary behavior. We conducted our study with three experts from the app team -- a software engineer, a project manager and a data scientist. We found in our study that most human-centered aspects were not originally considered when developing the first version of the application. We also report on other insights and challenges faced in RE for the health application, e.g., frequently changing requirements.

cs.SE↗

arXiv2023

Requirements Practices and Gaps When Engineering Human-Centered Artificial Intelligence Systems

[Context] Engineering Artificial Intelligence (AI) software is a relatively new area with many challenges, unknowns, and limited proven best practices. Big companies such as Google, Microsoft, and Apple have provided a suite of recent guidelines to assist engineering teams in building human-centered AI systems. [Objective] The practices currently adopted by practitioners for developing such systems, especially during Requirements Engineering (RE), are little studied and reported to date. [Method] This paper presents the results of a survey conducted to understand current industry practices in RE for AI (RE4AI) and to determine which key human-centered AI guidelines should be followed. Our survey is based on mapping existing industrial guidelines, best practices, and efforts in the literature. [Results] We surveyed 29 professionals and found most participants agreed that all the human-centered aspects we mapped should be addressed in RE. Further, we found that most participants were using UML or Microsoft Office to present requirements. [Conclusion] We identify that most of the tools currently used are not equipped to manage AI-based software, and the use of UML and Office may pose issues to the quality of requirements captured for AI. Also, all human-centered practices mapped from the guidelines should be included in RE.

cs.SE↗

arXiv2022

Requirements Engineering for Artificial Intelligence Systems: A Systematic Mapping Study

[Context] In traditional software systems, Requirements Engineering (RE) activities are well-established and researched. However, building Artificial Intelligence (AI) based software with limited or no insight into the system's inner workings poses significant new challenges to RE. Existing literature has focused on using AI to manage RE activities, with limited research on RE for AI (RE4AI). [Objective] This paper investigates current approaches for specifying requirements for AI systems, identifies available frameworks, methodologies, tools, and techniques used to model requirements, and finds existing challenges and limitations. [Method] We performed a systematic mapping study to find papers on current RE4AI approaches. We identified 43 primary studies and analysed the existing methodologies, models, tools, and techniques used to specify and model requirements in real-world scenarios. [Results] We found several challenges and limitations of existing RE4AI practices. The findings highlighted that current RE applications were not adequately adaptable for building AI systems and emphasised the need to provide new techniques and tools to support RE4AI. [Conclusion] Our results showed that most of the empirical studies on RE4AI focused on autonomous, self-driving vehicles and managing data requirements, and areas such as ethics, trust, and explainability need further research.

cs.SE↗

arXiv2022

Needs and Challenges of Personal Data Visualisations in Mobile Health Apps: User Survey

Personal data visualisations are becoming a critical contributor toward the successful adoption of mobile health (m-health) apps. Thus, understanding user needs and challenges when using mobile personal data visualisation is essential to ensuring the adoption of these apps. This paper presents the results of a user survey to understand users' demographics, tasks, needs, and challenges of using mobile personal data visualisations. We had 56 complete responses. The survey's key findings are: 1) 51\% of the users use multiple health tracking apps to achieve their goals/needs; 2) bar charts and pie charts are the most favourable charts to view health data; 3) users prefer to visualise their data using a mix of text and charts - explanation is essential. Furthermore, the top three challenges reported by the participants are: too much data displayed, overlapping text, and visualisations are not helpful in information exploration. On the other hand, users' top three encouragement factors are easy-to-read presented data, easy to navigate, and quality data are shown in the chart. Furthermore, fun and curiosity are the primary drivers of m-health tracking apps. Finally, based on survey results, we propose data visualisation designing and developing guidelines that should avoid the reported challenges and ensure user satisfaction. In future work, we plan to contextualise our study and investigate the pain and gain of data visualised in the following m-health domains: sports activities, heart monitoring, blood pressure, sleeping pattern, and eating habits.

cs.HC↗

arXiv2022

Analysis of Personal Data Visualization Reviews On mHealth Apps (short paper)

Mobile devices, specifically, smartphones proved easy and quick access to data visualisations throughout various tracking apps. Mobile health (mHealth) apps have given non-expert users access to data visualisation to track their activities and health-related issues such as heart tracking and medication. However, no work is done on user experience or perception of data visualisations in mHealth apps. App reviews offer an indirect anchor for researchers to examine how non-expert users perceive and interact with data visualisations and identify the key challenges and recommendations. This paper introduces an analysis of app reviews on data visualisations reported on a dataset of 250 mHealth apps on the Google Play Store. We identified 8,406 comments related to data visualisations. 919 neutral comments, 1,557 negative comments and 5,930 positive comments. From analysing the user reviews, functional requirements turned out to be the most common problem across these app reviews, followed by the look and feel and then data problems. A complete set of data visualisations seem to be the most well-received capability of mHealth apps. We used these comments to develop classification and data visualisation guidelines when developing mobile data visualisations.

cs.HC↗