arXiv subjects

Shing-Chi Cheung

Publications and source records attributed to Shing-Chi Cheung.

At least 19 recordsLinked to original sources

arXiv2026

GraphDroid: Asynchronous LLM-Based Mobile App GUI Testing via History-Aware Exploration and Hybrid Intent Fulfillment

Automated GUI testing is a widely adopted technique for ensuring mobile application quality by simulating user interactions to exercise functionalities. Despite the research breakthroughs in the past decades, covering complex functionalities that require multi-step action sequences still remains challenging. Traditional tools lack semantic understanding capability and can rarely synthesize such action sequences. Recent LLM-based tools can generate test intents describing target functionalities and leverage the LLM to fulfill the intents, but suffer from three key limitations: 1) loss of historical context for identifying uncovered functionalities, 2) synchronous intent generation that blocks exploration, and 3) per-step LLM-driven fulfillment incurring high cost and latency. To address these limitations, we propose GraphDroid, an intent-driven GUI testing framework that integrates a cluster-based memory mechanism to effectively identify uncovered functionalities from historically visited states for comprehensive application testing. For improving testing efficiency, GraphDroid adopts an asynchronous intent generation paradigm that eliminates the latency bottleneck and a hybrid intent fulfillment strategy that reserves the LLM for fulfilling complex intents while delegating simple intents to a lightweight heuristic algorithm. We evaluate GraphDroid on 41 real-world Android apps against six state-of-the-art baselines. Results show that GraphDroid outperforms all baselines, achieving up to 36.4% higher code coverage while incurring less than one eighth of the cost of the best pure LLM-based baseline. GraphDroid also exposes 19 bugs in the 41 apps and detects 13 of 52 crashes in the Themis bug benchmark, surpassing all the six baselines. Seven of the 19 bugs were previously unknown and we reported them to the developers. So far, four bugs have been confirmed and fixed.

cs.SE↗

arXiv2026

SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring

As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.

cs.CL↗

arXiv2026

Understanding Agent-Reactive Bugs at the Model-Harness Boundary: An Empirical Study of LLM Agent Issue Reports

LLM agents span command-line interfaces (e.g., Codex) and agent frameworks (e.g., LangChain), integrating backend LLMs with harness code that parses model outputs, controls agent loops, and manages context. Both the harness and LLM-generated responses jointly shape an agent's execution. This architecture gives rise to bugs that cannot be readily understood by inspecting either component alone, because some bugs occur only when a particular LLM response elicits an abnormal reaction from the agent. Prior empirical studies of agent bugs have largely attributed failures either to limited model capabilities or to harness-side defects, such as outdated APIs and configuration misalignment, without characterizing these AR bugs. We conduct the first empirical study focused on agent-reactive (AR) bugs. Through manual analysis of 255 bug reports from Codex, Gemini-CLI, LangChain, and CrewAI, we construct a two-axis taxonomy covering observable symptoms and the LLM behaviors that trigger them. Our findings show that many AR bugs manifest as silent errors without well-defined test oracles, which makes detection difficult. The stochasticity of LLM responses further complicates bug reproduction. We additionally examine fixes proposed by users and implemented by developers. This analysis exposes a mismatch: users frequently advocate harness-side guardrails, whereas developers may attribute the issue to the LLM or respond slowly to user-proposed fixes. These findings point to the need for mechanisms that help users and developers understand the root causes and resolutions of AR bugs. Overall, the study highlights challenges specific to LLM agents and motivates the design of test oracles, reproduction support, and fault-localization techniques for AR bugs.

cs.SE↗

arXiv2026

Inside the Skill Market: From Software Engineering Activities to Reusable Agent Skills

Software engineering (abbrev. SE) has continuously evolved through increasingly powerful forms of reuse, from source code and libraries to components and services. Recent advances in AI agents have introduced a potentially new reusable artifact: skills. Emerging agent skill repositories and marketplaces enable developers to package, share, and reuse SE expertise as reusable skills. This trend raises a fundamental question: what SE activities are being encapsulated into reusable skills? Existing studies primarily focus on a broad range of skills acquisition, safety, or benchmarking, while lacking a systematic understanding of SE-specific skills and their coverage across the software development lifecycle. To address this gap, we conduct the first large-scale empirical study of SE skills in public repositories and marketplaces. We collect and analyze a large corpus of SE skills, examining the activities they encapsulate, lifecycle coverage, evolution characteristics, and evaluation mechanisms. Our findings reveal that SE activities are increasingly becoming reusable artifacts via skills and suggest promising research opportunities for skill recommendation and engineering-oriented structuring, as well as the need for mechanisms to encapsulate high-context SE activities into reusable skills. Overall, our study provides the first activity-centric characterization of SE skills and reveals how SE activities are increasingly being transformed into reusable skills. These findings offer new insights into skill reuse, ecosystem development, and the future of agent-centric SE.

cs.SE↗

arXiv2026

Programming-by-Example for Batch-Editing Collision Meshes in 3D Software

As 3D software proliferates, software artifacts now extend beyond code and 2D user interfaces to include 3D assets. Among these assets, collision meshes are critical as they define the geometry used by physics engines for collision detection and physical interaction. Although existing tools can automatically generate collision meshes from visual meshes, they often fail to capture the intended interaction behavior. As a result, developers need to manually edit many heterogeneous collision meshes, a process that is time-consuming and challenging to scale. To address this problem, we present a neuro-symbolic program synthesis approach for batch-editing collision meshes. We formulate the task as a programming-by-example problem: given a family of collision meshes with the same editing intent and a small number of user demonstrations, our approach synthesizes a reusable program that captures the editing intent and applies it to non-demonstration meshes. We implement this in a tool named MeshForge, and evaluate it across 24 tasks on 600 collision meshes. MeshForge successfully synthesizes 23/24 tasks, requiring 2.2 demonstrations and 3.5 seconds of synthesis time on average.

cs.GR↗

arXiv2026

Coding Agents Are Guessing: Measuring Action-Boundary Violations in Underspecified DevOps Instructions

LLM coding agents are increasingly deployed to act autonomously on real production infrastructure. They execute shell commands, modify repositories, and call operational APIs. However, completing a task is not sufficient for safety. A wrong action can cause severe consequences. Existing agent benchmarks largely emphasize task completion, leaving open how agents behave under benign but underspecified instructions. We present UnderSpecBench, a benchmark for measuring action-boundary violations in coding agents (i.e., Claude Code, Codex, and OpenCode) on DevOps tasks. UnderSpecBench includes 69 task families grounded in documented incidents, CVEs, or tool behavior and organized across four DevOps capability domains and nine operational control surfaces. To isolate underspecification from task difficulty, each task keeps the same environment and ground-truth safe action while varying the instruction along three axes: intent clarity, target certainty, and blast radius. The resulting 2,208 prompt variants are evaluated with deterministic, side-effect-based oracles that separate Safe Success, Wrong Target, and OverScope outcomes; non-action runs are further classified as clarification, refusal, or deferment. Across five agent x model configurations using OpenCode, Claude Code, and Codex, the evaluation results show that underspecification does not mainly make agents fail; it makes them guess. 55.8-67.8% of runs violate at least one boundary. Target underspecification sharply degrades action quality, while blast-radius cues barely reduce action propensity. These findings show that completion-centric evaluation can overstate safe autonomy and motivate mitigations at the model, harness, and system layer.

cs.SE↗

arXiv2026

Cloak and Detonate: Scanner Evasion and Dynamic Detection of Agent Skill Malware

LLM coding agents increasingly rely on third-party agent skills from public marketplaces, which execute with the agent's privileges and create a software supply-chain attack surface: a malicious skill can steal credentials, exfiltrate source code, or install backdoors. Existing defenses use static skill scanners based on pattern matching or LLM-as-judge analysis, but it remains unclear whether they withstand adaptive evasions that preserve malicious behavior while changing payload appearance. This paper first presents an adversarial study of existing skill scanners through SkillCloak, a payload-preserving evasion framework that keeps the attack semantics intact while transforming their visible form. SkillCloak uses two complementary strategies: Structural Obfuscation, which rewrites visible payload indicators into semantically equivalent forms, and Self-Extracting Skill (SFS) Packing, which hides malicious components from the install-time view and restores them during agent execution. Across eight scanners and 1,613 in-the-wild malicious skills, SFS Packing bypasses every scanner at over 90%, while Structural Obfuscation bypasses over 80% on most static scanners and reaches 96% on a hybrid scanner, showing that appearance-based auditing is insufficient. Motivated by this finding, we propose SkillDetonate, a behavior-centric runtime auditor that executes skills in a sandbox and detects malicious effects through OS-boundary information-flow evidence rather than install-time appearance. SkillDetonate combines on-demand closure lift, which observes instructions materialized during execution, with marker-based taint analysis, which tracks sensitive-data flows across the agent context, files, processes, and network operations. The results show that SkillDetonate detects 97% of attacks at a 2% false-positive rate and sustains 87% detection on real-world malicious skills.

cs.CR↗

arXiv2026

Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training

Continual post-training aims to extend large language models (LLMs) with new knowledge, skills, and behaviors, yet it remains unclear when sequential updates enable capability transfer and when they cause catastrophic forgetting. Existing methods mitigate forgetting through sequential fine-tuning, replay, regularization, or model merging, but offer limited criteria for determining when incorporating new updates is beneficial or harmful. In this work, we study LLM continual post-training through three questions: What drives forgetting? When do sequentially acquired capabilities transfer or interfere? How can compatibility be used to control update integration? We address these questions through task geometry: we represent each post-training task by its parameter update and study the covariance geometry induced by the update. Our central finding is that: forgetting can be considered as a state-relative update-integration failure, it arises when the covariance geometries induced by tasks misalign with the geometry of the evolving model state. Sequential updates transfer when they remain compatible with the model state shaped by previous updates, and interfere when state-relative geometry conflict becomes high. Motivated by this finding, we propose Geometry-Conflict Wasserstein Merging (GCWM), a data-free update-integration method that constructs a shared Wasserstein metric via Gaussian Wasserstein barycenters and uses geometry conflict to gate geometry-aware correction. Across Qwen3 0.6B--14B on domain-continual and capability-continual settings, GCWM consistently outperforms data-free baselines, improving retention and final performance without replay data. These results identify geometry conflict as both an explanatory signal for forgetting and a practical control signal for LLM continual post-training.

cs.LG↗

arXiv2026

LiveFMBench: Unveiling the Power and Limits of Agentic Workflows in Specification Generation

Formal specification is essential for rigorous program verification, yet writing correct specifications remains costly and difficult to automate. Although large language models (LLMs) and agents have shown promising progress, their true capabilities and failure modes remain unclear. We present the first systematic and contamination-aware study of LLM- and agent-based formal specification generation for C programs. We introduce LiveFMBench, a continuously evolving benchmark of 630 ACSL (ANSI/ISO C Specification Language)-annotated C programs, including 360 newly collected cases designed to mitigate data leakage. Using this benchmark, we evaluate direct prompting with different sampling sizes, reasoning-enabled (thinking mode) inference, the agentic pipeline, and perform a fine-grained failure analysis. Experimental results reveal that naive evaluation substantially overestimates performance because models under direct prompting may exhibit unfaithful behaviors, such as deceiving automated provers or ignoring code-context constraints; after excluding such cases, the true specification generation accuracy drops by approximately 20\%. We further find that both increased sampling and thinking mode significantly improve success rates, with smaller models benefiting more from thinking mode. Agentic pipelines are particularly effective under low sampling budgets and on harder datasets. Failure analysis further shows that incorrect loop invariants are the dominant error type, while agentic pipelines notably reduce assertion errors. These results expose fundamental limitations in current LLM-based approaches and suggest they remain far from replacing human-authored formal specifications. We release LiveFMBench at https://huggingface.co/datasets/fm-universe/Live-FM-Bench and all evaluation artifacts to support future research.

cs.SE↗

arXiv2026

MR-Coupler: Automated Metamorphic Test Generation via Functional Coupling Analysis

Metamorphic testing (MT) is a widely recognized technique for alleviating the oracle problem in software testing. However, its adoption is hindered by the difficulty of constructing effective metamorphic relations (MRs), which often require domain-specific or hard-to-obtain knowledge. In this work, we propose a novel approach that leverages the functional coupling between methods, which is readily available in source code, to automatically construct MRs and generate metamorphic test cases (MTCs). Our technique, MR-Coupler, identifies functionally coupled method pairs, employs large language models to generate candidate MTCs, and validates them through test amplification and mutation analysis. In particular, we leverage three functional coupling features to avoid expensive enumeration of possible method pairs, and a novel validation mechanism to reduce false alarms. Our evaluation of MR-Coupler on 100 human-written MTCs and 50 real-world bugs shows that it generates valid MTCs for over 90% of tasks, improves valid MTC generation by 64.90%, and reduces false alarms by 36.56% compared to baselines. Furthermore, the MTCs generated by MR-Coupler detect 44% of the real bugs. Our results highlight the effectiveness of leveraging functional coupling for automated MR construction and the potential of MR-Coupler to facilitate the adoption of MT in practice. We also released the tool and experimental data to support future research.

cs.SE↗

arXiv2026

Demystifying the Silence of Correctness Bugs in PyTorch Compiler

Performance optimization of AI infrastructure is key to the fast adoption of large language models (LLMs). The PyTorch compiler (torch.compile), a core optimization tool for deep learning (DL) models (including LLMs), has received due attention. However, torch.compile is prone to correctness bugs, which cause incorrect outputs of compiled DL models without triggering exceptions, crashes, or warnings. These bugs pose a serious threat to the reliability of downstream LLM applications. Data from the PyTorch community shows that 19.2% of high-priority issues are incorrect outputs of compiled DL models induced by torch.compile bugs, the second-most-common bug category (only behind program crashes at 19.57%). However, no systematic study has been conducted to specifically characterize and thereby detect these bugs. In this paper, we present the first empirical study of the correctness bugs in torch.compile, examine their characteristics, and assess the effectiveness of existing fuzzers in detecting them. Based on our findings, we propose a proof-of-concept testing technique named AlignGuard, tailored specifically for detecting correctness bugs in torch.compile. AlignGuard incorporates bug characteristics distilled from our empirical study, applying LLM-based test mutation to existing test cases for correctness bug detection. At the time of writing, AlignGuard has successfully detected 23 new correctness bugs in recent torch.compile. All these bugs have been confirmed or fixed by the PyTorch development team, and over half (14/23) of them are even marked as high-priority bugs, underscoring the usefulness of our technique.

cs.SE↗

arXiv2026

Can Large Language Models Model Programs Formally?

In the digital age, ensuring the correctness, safety, and reliability of software through formal verification is paramount, particularly as software increasingly underpins critical infrastructure. Formal verification, split into theorem proving and model checking, provides a feasible and reliable path. Unlike theorem proving, which yields notable advances, model checking has been less focused due to the difficulty of automatic program modeling. To fill this gap, we introduce Model-Bench, a benchmark and an accompanying pipeline for evaluating and improving LLMs' program modeling capability by modeling Python programs into verification-ready model checking specifications checkable by its accompanying model checker. Model-Bench comprises 400 Python programs derived from three well-known benchmarks (HumanEval, MBPP, and LiveCodeBench). Our extensive experiments reveal significant limitations in LLMs' program modeling and further provide inspiring directions.

cs.SE↗

arXiv2026

SWE-ABS: Adversarial Benchmark Strengthening Exposes Inflated Success Rates on Test-based Benchmark

The SWE-Bench Verified leaderboard is approaching saturation, with the top system achieving 78.80%. However, we show that this performance is inflated. Our re-evaluation reveals that one in five "solved" patches from the top-30 agents are semantically incorrect, passing only because weak test suites fail to expose their errors. We present SWE-ABS, an adversarial framework that strengthens test suites through a two-stage pipeline: (1) coverage-driven augmentation using program slicing to target untested code regions, and (2) mutation-driven adversarial testing that synthesizes plausible but incorrect patches to expose semantic blind spots. On SWE-Bench Verified (500 instances), SWE-ABS strengthens 50.2% of instances, a 25.1x improvement over prior work, and rejects 19.71% of previously passing patches. As a result, the top agent's score decreases from 78.80% to 62.20%, leading to significant leaderboard reshuffling, with the previous top-ranked agent dropping to fifth place.

cs.SE↗

arXiv2026

ModelWisdom: An Integrated Toolkit for TLA+ Model Visualization, Digest and Repair

Model checking in TLA+ provides strong correctness guarantees, yet practitioners continue to face significant challenges in interpreting counterexamples, understanding large state-transition graphs, and repairing faulty models. These difficulties stem from the limited explainability of raw model-checker output and the substantial manual effort required to trace violations back to source specifications. Although the TLA+ Toolbox includes a state diagram viewer, it offers only a static, fully expanded graph without folding, color highlighting, or semantic explanations, which limits its scalability and interpretability. We present ModelWisdom, an interactive environment that uses visualization and large language models to make TLA+ model checking more interpretable and actionable. ModelWisdom offers: (i) Model Visualization, with colorized violation highlighting, click-through links from transitions to TLA+ code, and mapping between violating states and broken properties; (ii) Graph Optimization, including tree-based structuring and node/edge folding to manage large models; (iii) Model Digest, which summarizes and explains subgraphs via large language models (LLMs) and performs preprocessing and partial explanations; and (iv) Model Repair, which extracts error information and supports iterative debugging. Together, these capabilities turn raw model-checker output into an interactive, explainable workflow, improving understanding and reducing debugging effort for nontrivial TLA+ specifications. The website to ModelWisdom is available: https://model-wisdom.pages.dev. A demonstrative video can be found at https://www.youtube.com/watch?v=plyZo30VShA.

cs.SE↗

arXiv2026

CAM: A Causality-based Analysis Framework for Multi-Agent Code Generation Systems

Despite the remarkable success that Multi-Agent Code Generation Systems (MACGS) have achieved, the inherent complexity of multi-agent architectures produces substantial volumes of intermediate outputs. To date, the individual importance of these intermediate outputs to the system correctness remains opaque, which impedes targeted optimization of MACGS designs. To address this challenge, we propose CAM, the first \textbf{C}ausality-based \textbf{A}nalysis framework for \textbf{M}ACGS that systematically quantifies the contribution of different intermediate features for system correctness. By comprehensively categorizing intermediate outputs and systematically simulating realistic errors on intermediate features, we identify the important features for system correctness and aggregate their importance rankings. We conduct extensive empirical analysis on the identified importance rankings. Our analysis reveals intriguing findings: first, we uncover context-dependent features\textemdash features whose importance emerges mainly through interactions with other features, revealing that quality assurance for MACGS should incorporate cross-feature consistency checks; second, we reveal that hybrid backend MACGS with different backend LLMs assigned according to their relative strength achieves up to 7.3\% Pass@1 improvement, underscoring hybrid architectures as a promising direction for future MACGS design. We further demonstrate CAM's practical utility through two applications: (1) failure repair which achieves a 73.6\% success rate by optimizing top-3 importance-ranked features and (2) feature pruning that reduces up to 33.6\% intermediate token consumption while maintaining generation performance. Our work provides actionable insights for MACGS design and deployment, establishing causality analysis as a powerful approach for understanding and improving MACGS.

cs.SE↗

arXiv2025

Multi-Agent Systems for Dataset Adaptation in Software Engineering: Capabilities, Limitations, and Future Directions

Automating the adaptation of software engineering (SE) research artifacts across datasets is essential for scalability and reproducibility, yet it remains largely unstudied. Recent advances in large language model (LLM)-based multi-agent systems, such as GitHub Copilot's agent mode, promise to automate complex development workflows through coordinated reasoning, code generation, and tool interaction. This paper presents the first empirical study on how state-of-the-art multi-agent systems perform in dataset adaptation tasks. We evaluate Copilot, backed by GPT-4.1 and Claude Sonnet 4, on adapting SE research artifacts from benchmark repositories including ROCODE and LogHub2.0. Through a five-stage evaluation pipeline (file comprehension, code editing, command generation, validation, and final execution), we measure success rates, analyze failure patterns, and assess prompt-based interventions designed to enhance agent performance. Results show that current systems can identify key files and generate partial adaptations but rarely produce functionally correct implementations. Prompt-level interventions, especially providing execution error messages and reference code, substantially improve structural similarity to ground truth (from 7.25% to 67.14%), highlighting the importance of contextual and feedback-driven guidance. Our findings reveal both the promise and limitations of today's multi-agent LLM systems for dataset adaptation, and suggest concrete directions for building more reliable, self-correcting agents in future SE research.

cs.SE↗

arXiv2025

Optimization-Aware Test Generation for Deep Learning Compilers

Deep Learning (DL) compilers have been widely utilized to optimize DL models for efficient deployment across various hardware. Due to their vital role in the DL ecosystem, ensuring their reliability and security is critical. However, existing approaches have limitations in testing optimization stages, which is the core functionality of DL compilers, due to the difficulty in generating optimization-aware tests. In this paper, we proposed OATest, a novel approach for synthesizing optimization-aware computational graphs. The approach combines patterns extracted from documented tests for optimization and incorporates them into seed computational graphs, enabling broader exploration of optimization paths. To guarantee the optimization-awareness of generated graphs, OATest introduces the edges reusing strategy to establish strong connections between patterns and contexts. Additionally, to solve the validity challenge for the generated graphs, OATest employs an auxiliary layers addition strategy to resolve broken constraints. Equipped with two distinct test oracles, OATest applies differential testing to evaluate the two widely used DL compilers (i.e., TVM and ONNXRuntime). Our experimental results show that OATest outperforms the state-of-the-art method by detecting more bugs and achieving higher code coverage in TVM and ONNXRutimes. Additionally, OATest uncovers 58 previously unknown bugs, 36 of which have been confirmed or fixed by developers.

cs.SE↗

arXiv2025

LLM-Powered Detection of Price Manipulation in DeFi

Decentralized Finance (DeFi) smart contracts manage billions of dollars, making them a prime target for exploits. Price manipulation vulnerabilities, often via flash loans, are a devastating class of attacks causing significant financial losses. Existing detection methods are limited. Reactive approaches analyze attacks only after they occur, while proactive static analysis tools rely on rigid, predefined heuristics, limiting adaptability. Both depend on known attack patterns, failing to identify novel variants or comprehend complex economic logic. We propose PMDetector, a hybrid framework combining static analysis with Large Language Model (LLM)-based reasoning to proactively detect price manipulation vulnerabilities. Our approach uses a formal attack model and a three-stage pipeline. First, static taint analysis identifies potentially vulnerable code paths. Second, a two-stage LLM process filters paths by analyzing defenses and then simulates attacks to evaluate exploitability. Finally, a static analysis checker validates LLM results, retaining only high-risk paths and generating comprehensive vulnerability reports. To evaluate its effectiveness, we built a dataset of 73 real-world vulnerable and 288 benign DeFi protocols. Results show PMDetector achieves 88% precision and 90% recall with Gemini 2.5-flash, significantly outperforming state-of-the-art static analysis and LLM-based approaches. Auditing a vulnerability with PMDetector costs just $0.03 and takes 4.0 seconds with GPT-4.1, offering an efficient and cost-effective alternative to manual audits.

cs.CR↗