arXiv subjects

Christoph Csallner

Publications and source records attributed to Christoph Csallner.

15 recordsLinked to original sources

arXiv2026

Characterizing Visual Accessibility Issues in AI Developer Tools: An Empirical Study

AI-assisted developer tools increasingly mediate programming through chat panels, terminal agents, generated diffs, and streaming status output. These interaction surfaces may create visual accessibility barriers for blind, low-vision, and color-vision-deficient developers, yet little is known about how such barriers are reported in public tool ecosystems. We analyze issues and forum discussions from five AI developer tool ecosystems: GitHub Copilot in VS Code, Cursor, Claude Code, OpenAI Codex, and OpenCode. From 2,652 keyword-retrieved candidates, a three-model ensemble identified 600 unanimously positive visual accessibility reports. A stratified manual sanity check supported this conservative selection. Topic modeling and qualitative analysis identified three recurring categories: screen-reader and assistive-technology barriers; visual presentation, contrast, and differentiation problems; and readability, scaling, and control limitations in AI-specific interfaces. The relative prominence of these concerns varied across ecosystems and reflected differences in editor, terminal, chat, diff, and agent interaction surfaces. An exploratory metadata analysis further identified differences in reporter activity and, across the GitHub-based ecosystems, maintainer participation and closure processes. These findings show that the accessibility record of AI developer tools is shaped by both their interaction design and the reporting and maintenance practices of their surrounding ecosystems.

cs.SE↗

arXiv2026

ZIMPAF & RedPhuzz: High-fidelity Web Application Fuzzing via Branch, Language Construct, and Function Call Monitoring

We present ZIMPAF, runtime interpreter instrumentation, and RedPhuzz, a fuzzer, to address key limitations of state-of-the-art fuzzers: inefficient instrumentation, the lack of knowledge of the execution environment, and limited web domain knowledge. ZIMPAF implements a novel multi-granular runtime interpreter instrumentation that provides branch coverage, robust error and exception logging, function and language construct monitoring, and identification of user-supplied inputs used in branch instructions. The instrumentation is capable of identifying potentially vulnerable functions whose parameters are tainted, marking them as high-valued fuzzing targets, without performing taint analysis. It also employs a novel backward constant probe to infer potentially vulnerable functions whose parameters originate from constants, indicating their invulnerability and allowing them to be skipped. This information is utilized by RedPhuzz to perform highly-targeted function-and input-level fuzzing that goes beyond simple error-based fuzzing, but also detects silent vulnerabilities via multi-stage vulnerability detection. We also introduce three novel mutation strategies to achieve highly targeted and effective fuzzing: sanitization-aware, input-in-branch-aware, and data type-aware mutation. We evaluate RedPhuzz's performance with its predecessor (Phuzz) with 86 test cases across six benchmark web applications. RedPhuzz detects all vulnerabilities, while Phuzz fails to detect 16. RedPhuzz is 73% faster than Phuzz despite performing more tasks. ZIMPAF is faster than Phuzz's instrumentation (PCOV and UOPZ), while writing significantly more data. ZIMPAF achieves 2.1 to 41.22 times higher throughput than PCOV and UOPZ across five sampled benchmarks, and 0.87 times for one.

cs.CR↗

arXiv2026

Toward Inferring Accurate Context-free Grammars for Big Languages in a Black-box Setting

Black-box context-free grammar inference is crucial for program analysis, reverse engineering, program understanding, fuzzing, and security. But existing approaches such as Arvada, TreeVada, Kedavra, and Cucio struggle with scalability, accuracy, and grammar readability, especially on larger languages. To address this challenge, we introduce XVada with several new techniques for deterministic inference of context-free grammars. In an empirical comparison that avoids several pitfalls of recent studies, XVada improves on the highest-scoring competitor (TreeVada) both in grammar accuracy and grammar compactness. XVada also found a CVE in the widely used Python Liquid engine. Fuzzing based on the XVada-inferred grammar found five more bugs, which the Python Liquid developers fixed based on our bug reports. XVada and all experimental data and scripts are freely available.

cs.SE↗

arXiv2026

Do Autonomous Agents Contribute Test Code? A Study of Tests in Agentic Pull Requests

Testing is a critical practice for ensuring software correctness and long-term maintainability. As agentic coding tools increasingly submit pull requests (PRs), it becomes essential to understand how testing appears in these agent-driven workflows. Using the AIDev dataset, we present an empirical study of test inclusion in agentic pull requests. We examine how often tests are included, when they are introduced during the PR lifecycle and how test-containing PRs differ from non-test PRs in terms of size, turnaround time, and merge outcomes. Across agents, test-containing PRs are more common over time and tend to be larger and take longer to complete, while merge rates remain largely similar. We also observe variation across agents in both test adoption and the balance between test and production code within test PRs. Our findings provide a descriptive view of testing behavior in agentic pull requests and offer empirical grounding for future studies of autonomous software development.

cs.SE↗

arXiv2025

What Does It Take? Developing a Smartphone App that Motivates Older Adults to be Physically Active

Maintaining physical activity is essential for older adults' health and well-being, yet participation remains low. Traditional paper-based and in-person interventions have been effective but face scalability issues. Smartphone apps offer a potential solution, but their effectiveness in real-world use remains underexplored. Most prior studies take place in controlled environments, use specialized hardware, or rely on in-person training sessions or researcher-led setup. This study examines the feasibility and engagement of Senior Fit, a standalone mobile fitness app designed for older adults. We conducted continuous testing with 25 participants aged 65-85, refining the app based on their feedback to improve usability and accessibility. Our findings underscore both the potential and key challenges in designing digital health interventions. Older adults valued features such as video demonstrations and reminders that made activity feel accessible and motivating, yet some expressed frustration with manual logging and limited personalization. The Facebook group provided encouragement for some but excluded others unfamiliar with the platform. These results highlight the need for fitness apps that integrate flexible tracking, clear feedback, and low-barrier social support. We contribute design recommendations for creating inclusive mobile fitness tools that align with older adults' routines and capabilities, offering insights for future long-term, real-world deployments.

cs.HC↗

arXiv2025

Black-box Context-free Grammar Inference for Readable & Natural Grammars

Black-box context-free grammar inference is crucial for program analysis, reverse engineering, and security, yet existing tools such as Arvada, TreeVada, and Kedavra struggle with scalability, readability, and accuracy on large, complex languages. We present NatGI, a novel LLM-guided grammar inference framework that extends TreeVada's parse tree recovery with three key innovations: bracket-guided bubble exploration, LLM-driven bubble generation and non-terminal labeling, and hierarchical delta debugging (HDD) for systematic tree simplification. Bracket-guided exploration leverages syntactic cues such as parentheses to propose well-structured grammar fragments, while LLM guidance produces meaningful non-terminal names and selects more promising merges. Finally, HDD incrementally reduces unnecessary rules, which makes the grammars both compact and interpretable. In our experiments, we evaluate NatGI on a comprehensive benchmark suite ranging from small languages to larger ones such as lua, c, and mysql. Our results show that NatGI consistently outperforms strong baselines in terms of F1 score. On average, NatGI achieves an F1 score of 0.57, which is 25pp (percentage points) higher than the best-performing baseline, TreeVada. In the case of interpretability, our generated grammars perform significantly better than those produced by existing approaches. Leveraging LLM-based node renaming and bubble exploration, NatGI produces rules with meaningful non-terminal names and compact structures that align more closely with human intuition. As a result, developers and researchers can achieve higher accuracy while still being able to easily inspect, verify, and reason about the structure and semantics of the induced grammars.

cs.SE↗

arXiv2025

Early Accessibility: Automating Alt-Text Generation for UI Icons During App Development

Alt-text is essential for mobile app accessibility, yet UI icons often lack meaningful descriptions, limiting accessibility for screen reader users. Existing approaches either require extensive labeled datasets, struggle with partial UI contexts, or operate post-development, increasing technical debt. We first conduct a formative study to determine when and how developers prefer to generate icon alt-text. We then explore the ALTICON approach for generating alt-text for UI icons during development using two fine-tuned models: a text-only large language model that processes extracted UI metadata and a multi-modal model that jointly analyzes icon images and textual context. To improve accuracy, the method extracts relevant UI information from the DOM tree, retrieves in-icon text via OCR, and applies structured prompts for alt-text generation. Our empirical evaluation with the most closely related deep-learning and vision-language models shows that ALTICON generates alt-text that is of higher quality while not requiring a full-screen input.

cs.SE↗

arXiv2024

Inferring Alt-text For UI Icons With Large Language Models During App Development

Ensuring accessibility in mobile applications remains a significant challenge, particularly for visually impaired users who rely on screen readers. User interface icons are essential for navigation and interaction and often lack meaningful alt-text, creating barriers to effective use. Traditional deep learning approaches for generating alt-text require extensive datasets and struggle with the diversity and imbalance of icon types. More recent Vision Language Models (VLMs) require complete UI screens, which can be impractical during the iterative phases of app development. To address these issues, we introduce a novel method using Large Language Models (LLMs) to autonomously generate informative alt-text for mobile UI icons with partial UI data. By incorporating icon context, that include class, resource ID, bounds, OCR-detected text, and contextual information from parent and sibling nodes, we fine-tune an off-the-shelf LLM on a small dataset of approximately 1.4k icons, yielding IconDesc. In an empirical evaluation and a user study IconDesc demonstrates significant improvements in generating relevant alt-text. This ability makes IconDesc an invaluable tool for developers, aiding in the rapid iteration and enhancement of UI accessibility.

cs.HC↗

arXiv2023

Fast Deterministic Black-box Context-free Grammar Inference

Black-box context-free grammar inference is a hard problem as in many practical settings it only has access to a limited number of example programs. The state-of-the-art approach Arvada heuristically generalizes grammar rules starting from flat parse trees and is non-deterministic to explore different generalization sequences. We observe that many of Arvada's generalization steps violate common language concept nesting rules. We thus propose to pre-structure input programs along these nesting rules, apply learnt rules recursively, and make black-box context-free grammar inference deterministic. The resulting TreeVada yielded faster runtime and higher-quality grammars in an empirical comparison. The TreeVada source code, scripts, evaluation parameters, and training data are open-source and publicly available (https://doi.org/10.6084/m9.figshare.23907738).

cs.SE↗

arXiv2023

Replicability Study: Corpora For Understanding Simulink Models & Projects

Background: Empirical studies on widely used model-based development tools such as MATLAB/Simulink are limited despite the tools' importance in various industries. Aims: The aim of this paper is to investigate the reproducibility of previous empirical studies that used Simulink model corpora and to evaluate the generalizability of their results to a newer and larger corpus, including a comparison with proprietary models. Method: The study reviews methodologies and data sources employed in prior Simulink model studies and replicates the previous analysis using SLNET. In addition, we propose a heuristic for determining code-generating Simulink models and assess the open-source models' similarity to proprietary models. Results: Our analysis of SLNET confirms and contradicts earlier findings and highlights its potential as a valuable resource for model-based development research. We found that open-source Simulink models follow good modeling practices and contain models comparable in size and properties to proprietary models. We also collected and distribute 208 git repositories with over 9k commits, facilitating studies on model evolution. Conclusions: The replication study offers actionable insights and lessons learned from the reproduction process, including valuable information on the generalizability of research findings based on earlier open-source corpora to the newer and larger SLNET corpus. The study sheds light on noteworthy attributes of SLNET, which is self-contained and redistributable.

cs.SE↗

arXiv2023

Searching Mobile App Screens via Text + Doodle

Locating a specific mobile application screen from existing repositories is restricted to basic keyword searches, such as Google Image Search, or necessitates a complete query screen image, as in the case of Swire. However, interactive partial sketch-based solutions like PSDoodle have limitations, including inaccuracy and an inability to consider text appearing on the screen. A potentially effective solution involves implementing a system that provides interactive partial sketching functionality for efficiently structuring user interface elements. Additionally, the system should incorporate text queries to enhance its capabilities further. Our approach, TpD, represents the pioneering effort to enable an iterative search of screens by combining interactive sketching and keyword search techniques. TpD is built on a combination of the Rico repository of approximately 58k Android app screens and the PSDoodle. Our evaluation with third-party software developers showed that PSDoodle provided higher top-10 screen retrieval accuracy than state-of-the-art Swire and required less time to complete a query than other interactive solutions.

cs.IR↗

arXiv2022

PSDoodle: Searching for App Screens via Interactive Sketching

Keyword-based mobile screen search does not account for screen content and fails to operate as a universal tool for all levels of users. Visual searching (e.g., image, sketch) is structured and easy to adopt. Current visual search approaches count on a complete screen and are therefore slow and tedious. PSDoodle employs a deep neural network to recognize partial screen element drawings instantly on a digital drawing interface and shows results in real-time. PSDoodle is the first tool that utilizes partial sketches and searches for screens in an interactive iterative way. PSDoodle supports different drawing styles and retrieves search results that are relevant to the user's sketch query. A short video demonstration is available online at: https://youtu.be/3cVLHFm5pY4

cs.CV↗

arXiv2022

PSDoodle: Fast App Screen Search via Partial Screen Doodle

Searching through existing repositories for a specific mobile app screen design is currently either slow or tedious. Such searches are either limited to basic keyword searches (Google Image Search) or require as input a complete query screen image (SWIRE). A promising alternative is interactive partial sketching, which is more structured than keyword search and faster than complete-screen queries. PSDoodle is the first system to allow interactive search of screens via interactive sketching. PSDoodle is built on top of a combination of the Rico repository of some 58k Android app screens, the Google QuickDraw dataset of icon-level doodles, and DoodleUINet, a curated corpus of some 10k app icon doodles collected from hundreds of individuals. In our evaluation with third-party software developers, PSDoodle provided similar top-10 screen retrieval accuracy as the state of the art from the SWIRE line of work, while cutting the average time required about in half.

cs.SE↗

arXiv2022

SLNET: A Redistributable Corpus of 3rd-party Simulink Models

MATLAB/Simulink is widely used for model-based design. Engineers create Simulink models and compile them to embedded code, often to control safety-critical cyber-physical systems in automotive, aerospace, and healthcare applications. Despite Simulink's importance, there are few large-scale empirical Simulink studies, perhaps because there is no large readily available corpus of third-party open-source Simulink models. To enable empirical Simulink studies, this paper introduces SLNET, the largest corpus of freely available third-party Simulink models. SLNET has several advantages over earlier collections. Specifically, SLNET is 8 times larger than the largest previous corpus of Simulink models, includes fine-grained metadata, is constructed automatically, is self-contained, and allows redistribution. SLNET is available under permissive open-source licenses and contains all of its collection and analysis tools.

cs.SE↗

arXiv2022

SLGPT: Using Transfer Learning to Directly Generate Simulink Model Files and Find Bugs in the Simulink Toolchain

Finding bugs in a commercial cyber-physical system (CPS) development tool such as Simulink is hard as its codebase contains millions of lines of code and complete formal language specifications are not available. While deep learning techniques promise to learn such language specifications from sample models, deep learning needs a large number of training data to work well. SLGPT addresses this problem by using transfer learning to leverage the powerful Generative Pre-trained Transformer 2 (GPT-2) model, which has been pre-trained on a large set of training data. SLGPT adapts GPT-2 to Simulink with both randomly generated models and models mined from open-source repositories. SLGPT produced Simulink models that are both more similar to open-source models than its closest competitor, DeepFuzzSL, and found a super-set of the Simulink development toolchain bugs found by DeepFuzzSL.

cs.SE↗