Ali Mesbah

Noor Nashid, Daniel Ding, Keheliya Gallaba et al.

Automated program repair has traditionally focused on single-hunk defects, overlooking multi-hunk bugs that are prevalent in real-world systems. Repairing these bugs requires coordinated edits across multiple, disjoint code regions, posing substantially greater challenges. We present the first systematic study of LLM-driven coding agents (Claude Code, Codex, Gemini-cli, and Qwen Code) on this task. We evaluate these four state-of-the-art agents on 404 multi-hunk bugs from the PolyHunk dataset, yielding 1,616 repair trajectories for large-scale behavioral analysis. We employ fine-grained metrics to assess localization, repair accuracy, regression behavior, and operational dynamics across agents. We find that localization capability varies substantially, with Codex achieving the highest success rate (75.3%) and Qwen Code the lowest (40.4%). Repair accuracy also differs widely, ranging from 26.98% (Qwen Code) to 92.82% (Claude Code), and consistently declines with increasing bug dispersion and complexity (hunk divergence and spatial proximity). High-performing agents (Claude Code and Codex) demonstrate superior semantic consistency, achieving positive average regression reduction, whereas lower-performing agents often introduce new test failures. Notably, agents do not fail fast; failed repairs consume substantially more resources (33%-440% more input tokens) and require longer execution time (35%-330%). Additionally, we developed Maple to provide agents with repository-level context. Empirical results show that Maple improves repair accuracy of Gemini-cli by ~21% through enhanced localization. By analyzing fine-grained metrics and trajectory-level analysis, this study moves beyond accuracy to explain how coding agents localize, reason, and act during multi-hunk repair. Our findings underscore the impact of bug divergence and spatial proximity on multi-hunk repair success for coding agents.

Mobina Shahbandeh, Parsa Alian, Noor Nashid et al.

End-to-end web testing is challenging due to the need to explore diverse web application functionalities. Current state-of-the-art methods, such as WebCanvas, are not designed for broad functionality exploration; they rely on specific, detailed task descriptions, limiting their adaptability in dynamic web environments. We introduce NaviQAte, which frames web application exploration as a question-and-answer task, generating action sequences for functionalities without requiring detailed parameters. Our three-phase approach utilizes advanced large language models like GPT-4o for complex decision-making and cost-effective models, such as GPT-4o mini, for simpler tasks. NaviQAte focuses on functionality-guided web application navigation, integrating multi-modal inputs such as text and images to enhance contextual understanding. Evaluations on the Mind2Web-Live and Mind2Web-Live-Abstracted datasets show that NaviQAte achieves a 44.23% success rate in user task navigation and a 38.46% success rate in functionality navigation, representing a 15% and 33% improvement over WebCanvas. These results underscore the effectiveness of our approach in advancing automated web application testing.

Joshua Hang Sai Ip, Georgios Makrygiorgos, Ali Mesbah

Bayesian Optimization (BO) is a principled approach for optimizing expensive black-box functions, with Expected Improvement (EI) being one of the most widely used acquisition functions. Despite its empirical success, EI is known to be overly exploitative and can converge to suboptimal stationary points. We propose Expected Improvement via Gradient Norms (EI-GN), a novel acquisition function that applies the improvement principle to a gradient-aware auxiliary objective, thereby promoting sampling in regions that are both high-performing and approaching first-order stationarity. EI-GN relies on gradient observations used to learn gradient-enhanced surrogate models that enable principled gradient inference from function evaluations. We derive a tractable closed-form expression for EI-GN that allows efficient optimization and show that the proposed acquisition is consistent with the improvement-based acquisition framework. Empirical evaluations on standard BO benchmarks demonstrate that EI-GN yields consistent improvements against standard baselines. We further demonstrate applicability of EI-GN to control policy learning problems.

Joshua Hang Sai Ip, Georgios Makrygiorgos, Ali Mesbah

Deep neural networks are increasingly used as an effective parameterization of control policies in various learning-based control paradigms. For continuous-time optimal control problems (OCPs), which are central to many decision-making tasks, control policy learning can be cast as a neural ordinary differential equation (NODE) problem wherein state and control constraints are naturally accommodated. This paper presents a NODE approach to solving continuous-time OCPs for the case of stabilizing a known constrained nonlinear system around a target state. The approach, termed Lyapunov-NODE control (L-NODEC), uses a novel Lyapunov loss formulation that incorporates an exponentially-stabilizing control Lyapunov function to learn a state-feedback neural control policy, bridging the gap of solving continuous-time OCPs via NODEs with stability guarantees. The proposed Lyapunov loss allows L-NODEC to guarantee exponential stability of the controlled system, as well as its adversarial robustness to perturbations to the initial state. The performance of L-NODEC is illustrated in two problems, including a dose delivery problem in plasma medicine. In both cases, L-NODEC effectively stabilizes the controlled system around the target state despite perturbations to the initial state and reduces the inference time necessary to reach the target.

Mohammad Bajammal, Ali Mesbah

Page segmentation is a web page analysis process that divides a page into cohesive segments, such as sidebars, headers, and footers. Current page segmentation approaches use either the DOM, textual content, or rendering style information of the page. However, these approaches have a number of drawbacks, such as a large number of parameters and rigid assumptions about the page, which negatively impact their segmentation accuracy. We propose a novel page segmentation approach based on visual analysis of localized adjacency regions. It combines DOM attributes and visual analysis to build features of a given page and guide an unsupervised clustering. We evaluate our approach on 35 real-world web pages, and examine the effectiveness and efficiency of segmentation. The results show that, compared with state-of-the-art, our approach achieves an average of 156% increase in precision and 249% improvement in F-measure.

Davood Mazinanian, Mohammad Bajammal, Ali Mesbah

A wide range of analysis and testing techniques targeting modern web apps rely on the automated exploration of their state space by firing events that mimic user interactions. However, finding out which elements are actionable in web apps is not a trivial task. To improve the efficacy of exploring the event space of web apps, we propose a browser-independent, instrumentation-free approach based on structural and visual stylistic cues. Our approach, implemented in a tool called StyleX, employs machine learning models, trained on 700,000 web elements from 1,000 real-world websites, to predict actionable elements on a webpage a priori. In addition, our approach uses stylistic cues for ranking these actionable elements while exploring the app. Our actionable predictor models achieve 90.14\% precision and 87.76\% recall when considering the click event listener, and on average, 75.42\% precision and 77.76\% recall when considering the five most-frequent event types. Our evaluations show that StyleX can improve the JavaScript code coverage achieved by a general-purpose crawler by up to 23\%.

Marjane Namavar, Noor Nashid, Ali Mesbah

Training a deep learning model on source code has gained significant traction recently. Since such models reason about vectors of numbers, source code needs to be converted to a code representation before vectorization. Numerous approaches have been proposed to represent source code, from sequences of tokens to abstract syntax trees. However, there is no systematic study to understand the effect of code representation on learning performance. Through a controlled experiment, we examine the impact of various code representations on model accuracy and usefulness in deep learning-based program repair. We train 21 different generative models that suggest fixes for name-based bugs, including 14 different homogeneous code representations, four mixed representations for the buggy and fixed code, and three different embeddings. We assess if fix suggestions produced by the model in various code representations are automatically patchable, meaning they can be transformed to a valid code that is ready to be applied to the buggy code to fix it. We also conduct a developer study to qualitatively evaluate the usefulness of inferred fixes in different code representations. Our results highlight the importance of code representation and its impact on learning and usefulness. Our findings indicate that (1) while code abstractions help the learning process, they can adversely impact the usefulness of inferred fixes from a developer's point of view; this emphasizes the need to look at the patches generated from the practitioner's perspective, which is often neglected in the literature, (2) mixed representations can outperform homogeneous code representations, (3) bug type can affect the effectiveness of different code representations; although current techniques use a single code representation for all bug types, there is no single best code representation applicable to all bug types.

Rahulkrishna Yandrapally, Ali Mesbah

Automated model-based test generation presents a viable alternative to the costly manual test creation currently employed for regression testing of web apps. However, existing model inference techniques rely on threshold-based whole-page comparison to establish state equivalence, which cannot reliably identify near-duplicate web pages in modern web apps. Consequently, existing techniques produce inadequate models for dynamic web apps, and fragile test oracles, rendering the generated regression test suites ineffective. We propose a model-based test generation technique, FRAGGEN, that eliminates the need for thresholds, by employing a novel state abstraction based on page fragmentation to establish state equivalence. FRAGGEN also uses fine-grained page fragment analysis to diversify state exploration and generate reliable test oracles. Our evaluation shows that FRAGGEN outperforms existing whole-page techniques by detecting more near-duplicates, inferring better web app models and generating test suites that are better suited for regression testing. On a dataset of 86,165 state-pairs, FRAGGEN detected 123% more near-duplicates on average compared to whole-page techniques. The crawl models inferred by FRAGGEN have 62% more precision and 70% more recall on average. FRAGGEN also generates reliable regression test suites with test actions that have nearly 100% success rate on the same version of the web app even if the execution environment is varied. The test oracles generated by FRAGGEN can detect 98.7% of the visible changes in web pages while being highly robust, making them suitable for regression testing.