arXiv subjects

Aldeida Aleti

Publications and source records attributed to Aldeida Aleti.

At least 19 recordsLinked to original sources

arXiv2026

Detecting Incorrect LLM Explanations for Automated Program Repair

LLM-generated explanations help developers understand automated program repair (APR) rationale and assess patch quality. Incorrect explanations can mislead developers about a repair's reliability, making their detection important. However, manually inspecting the growing volume of generated repairs is impractical, motivating automated detection. Existing methods mainly target incorrect answers from general question-answering LLMs, often by sampling multiple responses and detecting contradictions. Applied to APR, they face two limitations: (1) they may depend on external verifiers that are unavailable or lack APR-specific reasoning; and (2) sampling strategies such as paraphrasing or re-prompting the same input often yield superficial, redundant responses that repeat the original error rather than expose it. We propose MetaReJ, the first method designed to detect incorrect APR explanations using metamorphic testing. Inspired by debate and viva examinations, MetaReJ challenges explanations through an argumentative strategy without requiring external verifiers. It applies novel metamorphic relations (MRs) to recursively question the LLM under equivalent and non-equivalent repair scenarios and detect contradictions. To obtain diverse samples, MetaReJ transforms the buggy program into variants that preserve or alter the underlying bug, then asks the LLM how to repair each variant. Contradictions among responses are encoded as MR violations, which indicate that the original explanation is incorrect. We evaluate MetaReJ on a manually labelled dataset of 1,756 explanations from multiple LLM-based APR systems. MetaReJ achieves an MCC of up to 0.43, outperforming prior methods by up to 333.3% in MCC.

cs.SE↗

arXiv2026

Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics

The vocabulary of a language model is its native semantics, and the embedding and unembedding fix the meaning of a state at the two ends of the residual stream. Between them, existing methods reuse a fixed lens or treat states as a point cloud, and neither reads a state as the model does. We argue that every hidden state has a native reference frame, induced by the layers that remain and adaptive to the context. Lexical reference frames approximate it in a model-native and context-aware way, reading a state through the remaining layers with a given number of rounds of attention between positions and a typical attention output for the rest, fitting nothing to the model's output. No lens with fixed directions per layer can reach a mean cosine with the native frame's lexical readouts above the square root of their coherence, 0.30 and 0.35 at mid-depth of SmolLM3-3B and OLMo3-7B. LexFrame[1], of order 1, reaches 0.97 and 0.97, and its readings are on average within 0.16 and 0.15 nats of the model's output. Its coefficient differences attribute predictions more faithfully than attribution patching, and its lexical basis steers predictions while sparing other tokens. On average, its readings settle on the model's top token over the final 0.82 and 0.81 of the layers, those of lenses over at most 0.11. These frames thus make every hidden state readable, attributable and steerable.

cs.CL↗

arXiv2026

Mind the Gaps: From Failure Attribution to Closed-Form Repair of Code Language Models

Code language models must be maintained like the software around them: when a library evolves, a model keeps writing the interface that it saw during training. Repairing the model itself lets one correction reach all downstream uses. Existing repair methods attribute a failure to neurons, select the highest-ranked ones, and apply a generic update. This pipeline assumes that the attributed neurons are the ones to patch and that a generic update fits every failure, and neither assumption has been examined. We examine both on executable API evolution tasks in Python and Rust with three code models and identify two gaps. The targeting gap separates the neurons that failure attribution targets from the neurons that can carry a patch: their top sets have a Jaccard overlap of only 0.15 to 0.20. The tailoring gap separates a generic update from a patch built for the failure: the patches that different carrier neurons need are nearly orthogonal. To address both gaps, we propose ASTRA. It targets neurons by contrastive semantics, an attribution that scores a neuron by its contribution to the logit contrast between the target token and the produced token. It then tailors the patch by solving one small linear system in closed form, which corrects all failing tokens of a sample jointly and needs neither an optimizer nor a backward pass. Contrastive semantics selects significantly better carrier neurons than gradient-based attributions in 3 of 6 settings and comparable ones in the others. On average, ASTRA reaches 66.7 percent Pass@1, against 48.4 percent for the best of AlphaEdit, STAR and low-rank adaptation, and repairs a sample in 2.7 seconds. It is the best method in all 6 settings and for every type of API change, and this advantage persists under an unseen phrasing of the test prompt. Its side effects on unrelated code are small on the large models and larger on the small one.

cs.SE↗

arXiv2026

Requirements-Augmented Generation for Trustworthy Acceptance Testing of LLM-Based Software

LLM-based software (LBS) integrates large language models as core components to deliver flexible, personalised responses. Unlike traditional software with deterministic outputs, LBSs exhibit context-dependent, stochastic behaviour that renders classical acceptance testing and test oracles insufficient: the same query may require fundamentally different responses depending on user personas and software context. This gap creates an urgent need for automated acceptance testing frameworks that autonomously interpret user instructions, while reliably inferring user intentions in a changing environment. In this paper, we present an automated acceptance testing framework for LBS with calibrated verdict reliability via two technical contributions. First, we introduce Requirements-Augmented Generation (REAG), which interprets user intentions by retrieving relevant software requirements, domain knowledge, and personas via adaptive RAG and self-reasoning to generate context-aware test oracles. Second, recognising that oracle generation may retrieve irrelevant constraints, misinterpret intent, or hallucinate requirements, we introduce a confidence-calibrated cascade judgment. This method quantifies verdict reliability via simulated expert agreement -- accepting high-confidence verdicts, escalating ambiguous cases, or abstaining when uncertain -- with empirical reliability guarantees backed by conformal risk control. An industrial case study on a production nutrition advisory application demonstrates that REAG achieves a 3.91/5 oracle quality score, reaching qualified or marginal oracle quality in 82% of cases. The confidence-calibrated cascade achieves 98.8% accuracy, improves oracle quality from 3.91 to 4.30 by filtering unqualified outputs, and delivers a 31.7% cost-efficiency improvement over single-judge baselines, validating industrial viability

cs.SE↗

arXiv2026

CodeChat-Eval: Evaluating Large Language Models in Multi-Turn Code Refinement Dialogues

Large Language Models (LLMs) are increasingly used in software engineering to generate and refine code. In practice, developers often continue from an initial code generation request with follow-up refinement instructions, such as requests to improve style, restructure implementation, or change the execution strategy while preserving the intended behaviour. However, existing benchmarks generally omit this multi-turn code refinement dialogue setting and therefore cannot evaluate whether LLMs maintain functional correctness, i.e., whether the refined code still passes the test suite for the original task. To address this limitation, we introduce CodeChat-Eval, an evaluation framework that constructs evaluation sessions from multi-turn code refinement dialogues using a dynamic instruction selection algorithm. Our empirical study on open-weight and proprietary LLMs observes a statistically significant decrease ranging from 19.2% (GPT-5 Nano) to 69.2% (Llama 3.1 8B) in functional correctness over multi-turn refinement. The largest correctness drops are associated with logic-level refinements and additive change requests. These findings indicate that LLMs struggle to maintain functional correctness during multi-turn code refinement dialogues, and highlight the need for benchmarks that evaluate functionality-preserving refinement beyond single-turn generation.

cs.SE↗

arXiv2026

SemRF: A Semantic Reference Frame for Residual-Stream Dynamics in Language Models

Residual-stream analysis asks how language-model computation evolves across depth, but intermediate decoding requires comparable readout coordinates across layers. If embedding anchors and unembedding readout disagree on the chosen span, apparent motion may reflect measurement drift rather than computation. We introduce \emph{Semantic Reference Frames} (SemRF), an anchor-based formalism separating semantic measurement from residual dynamics. A SemRF fixes anchors and measures states against them. Pseudo-inverse tying gives exact synchronization; under restricted bi-invertibility, SemRF yields stable semantic-basis coordinates, distortion bounds, and near-identity changes. With the frame fixed, residual computation becomes a depthwise semantic trajectory. The anchors induce a semantic Voronoi diagram: distance, or evidence such as logits, assigns each layer to a coarse cell, while coordinates retain within-cell motion and margins. We define layerwise steps, contribution profiles, and imbalance diagnostics, then use the Voronoi trace to define a margin-relaxed tube. The canonical trace is the minimum-action path inside this tube; when nonempty with positive quadratic weight, it is unique and obeys a discrete spline equation away from active constraints. Excess action controls step, curvature, and profile mismatch. Low curvature implies piecewise-linear compressibility and local knowledge density: lower trace complexity means fewer semantic knots. Through the parameter-to-trajectory map, this gives a conditional link to parameter efficiency: among admissible settings fitting data, lower-action and lower-complexity traces use fewer semantic degrees of freedom. The guarantees require controlled interface error and small projection residual under explicit tube constraints.

cs.LG↗

arXiv2026

MORTAR: Multi-turn Metamorphic Testing for LLM-based Dialogue Systems

With the widespread application of LLM-based dialogue systems in daily life, quality assurance has become more important than ever. Recent research has successfully introduced methods to identify unexpected behaviour in single-turn testing scenarios. However, multi-turn interaction is the common real-world usage of dialogue systems, yet testing methods for such interactions remain underexplored. This is largely due to the oracle problem in multi-turn testing, which continues to pose a significant challenge for dialogue system developers and researchers. In this paper, we propose MORTAR, a metamorphic multi-turn dialogue testing approach, which mitigates the test oracle problem in testing LLM-based dialogue systems. MORTAR formalises the multi-turn testing for dialogue systems, and automates the generation of question-answer dialogue test cases with multiple dialogue-level perturbations and metamorphic relations (MRs). The automated MR matching mechanism allows MORTAR more flexibility and efficiency in metamorphic testing. The proposed approach is fully automated without reliance on LLM judges. In testing six popular LLM-based dialogue systems, MORTAR reaches significantly better effectiveness with over 150\% more bugs revealed per test case when compared to the single-turn metamorphic testing baseline. Regarding the quality of bugs, MORTAR reveals higher-quality bugs in terms of diversity, precision and uniqueness. MORTAR is expected to inspire more multi-turn testing approaches, and assist developers in evaluating the dialogue system performance more comprehensively with constrained test resources and budget.

cs.SE↗

arXiv2026

Trustworthy AI Software Engineers

With the rapid rise of AI coding agents, the fundamental premise of what it means to be a software engineer is in question. In this vision paper, we examine what it means for an AI agent to be considered a software engineer and then critically think about what makes such an agent trustworthy. Grounded in established definitions of SE (SE) and informed by recent research on agentic AI systems, we conceptualise AI software engineers as participants in human-AI SE teams composed of human software engineers and AI agents, and we distinguish trustworthiness as a key property of these systems and actors rather than a subjective human attitude. Extending on historical perspectives and emerging visions, we identify key dimensions that contribute to the trustworthiness of AI software engineers, spanning technical quality, transparency and accountability, epistemic humility, and societal and ethical alignment. Beyond defining these dimensions, we address a critical but underexplored challenge: how trustworthiness can be operationalised in practice. We therefore introduce the notion of evidence-centric inspection, arguing that developers should evaluate selective signals and justifications of trustworthiness rather than raw outputs, and we outline implications for rethinking verification, validation, and code review in human-AI SE teams.

cs.SE↗

arXiv2026

Beyond Neural Incompatibility: Cross-Scale Knowledge Transfer in Language Models through Latent Semantic Alignment

Language Models (LMs) encode substantial knowledge in their parameters, yet it remains unclear how to transfer such knowledge in a fine-grained manner, namely parametric knowledge transfer (PKT). A central challenge is to make cross-scale transfer effective and efficient when source and target models differ in architecture and parameterization, making direct parameter reuse strongly limited by neural incompatibility. In this paper, we identify latent semantic alignment as the key prerequisite for cross-scale knowledge transfer. Instead of directly moving layer parameters, our approach uses activations as the transfer medium. \textsc{SemAlign} has two stages: an \emph{layer attribution} stage that attributes task-relevant source layers and selects exactly one source layer for each target layer, and a \emph{semantic alignment} stage that pairs them layer by layer and optimizes the target with source-side semantic supervision. The alignment is carried out in latent space through semantic decomposition and recomposition. During the shallow-to-deep transfer, only the frontier target layer is trainable. The layer objective supervises the residual contribution of that layer by matching centered token-token relation geometry against an aligned supervisory residual, while output KL preserves source-level predictive behavior. The transferred medium is therefore neither a parameter block nor an absolute hidden state, but target-space residual geometry induced by paired source-layer supervision. Evaluations on four benchmarks demonstrate the efficacy of \textsc{SemAlign}, and further analysis confirms that semantic decomposition and recomposition provide a stable mechanism for cross-scale knowledge transfer.

cs.CL↗

arXiv2026

UntrustVul: An Automated Approach for Identifying Untrustworthy Alerts in Vulnerability Detection Models

Machine learning (ML) has shown promise in vulnerability detection, but ML detectors may rely on irrelevant code features, causing them to highlight non-vulnerable lines as suspicious. Such misleading predictions increase developers' manual effort and may lead to incorrect patching strategies, motivating the need to identify untrustworthy predictions automatically. We present UntrustVul, an approach for detecting untrustworthy vulnerability predictions by identifying suspicious lines that are inherently unrelated to vulnerabilities. UntrustVul leverages patterns from historical vulnerable lines and flags predictions as untrustworthy when the highlighted lines neither match known vulnerability patterns nor influence lines that do. A line is considered vulnerability-irrelevant if it does not resemble historical vulnerabilities and all its successors in the data and control dependency graph are also vulnerability-irrelevant. The approach is designed conservatively to minimise misclassifying trustworthy predictions as untrustworthy. We evaluate UntrustVul on 115K predictions from four models across the BigVul, MegaVul, SARD, and PrimeVul datasets. Results show that UntrustVul achieves AUC scores of 70%-88% and F1-scores of 82%-94%, outperforming existing approaches by 6%-59% in AUC and 13%-92% in F1-score.

cs.SE↗

arXiv2026

Rethinking Weight Tying: Pseudo-Inverse Tying for LM Stable Training and Updates

Weight tying is widely used in compact language models to reduce parameters by sharing the token table between the input embedding and the output projection. However, parameter sharing alone does not guarantee a stable token interface: during training, the correspondence between encoding tokens into hidden states and decoding hidden states into logits can drift, worsening optimization sensitivity and weakening explainability probes that rely on a meaningful vocabulary-space decoder. We propose Pseudo-Inverse Tying (PIT), which synchronizes embedding and unembedding as coupled projections of a shared latent token memory, guaranteeing a pseudo-inverse-consistent interface throughout training. PIT maintains an orthonormal shared memory, obtained by polar initialization from a source checkpoint for continued pretraining or by random orthonormal initialization for from-scratch pretraining, and introduces a learned symmetric positive definite hidden-space transform parameterized via a Cholesky factor. The output head applies this transform to hidden states before the vocabulary projection, while the embedding applies the inverse transform to token vectors using stable triangular solves, avoiding explicit pseudo-inverse recomputation and vocabulary-sized auxiliary parameters. Beyond improving training stability, PIT provides a cleaner substrate for logit-lens-style and vocabulary-space explainability probes by keeping the input and output token geometries synchronized. We evaluate PIT on on-device models spanning 256M-1.3B parameters. The results show that PIT improves continued-pretraining stability, enforces near-exact token-interface consistency across settings, and yields more predictable lightweight adaptation after continued pretraining, while from-scratch pretraining reveals a trade-off between strict interface consistency and unconstrained optimization.

cs.CL↗

arXiv2026

ViBR: Automated Bug Replay from Video-based Reports using Vision-Language Models

Bug reports play a critical role in software maintenance by helping users convey encountered issues to developers. Recently, GUI screen capture videos have gained popularity as a bug reporting artifact due to their ease of use and ability to retain rich contextual information. However, automatically reproducing bugs from such recordings remains a significant challenge. Existing methods often rely on fragile image-processing heuristics, explicit touch indicators, or pre-constructed UI transition graphs, which require non-trivial instrumentation and app-specific setup. This paper presents ViBR, a lightweight and fully automated approach that reproduces bugs directly from GUI recordings. Specifically, ViBR combines CLIP-based embedding similarity for action boundary segmentation with Vision-Language Models (VLMs) for region-aware GUI state comparison and guided bug replay. Experimental results show that ViBR successfully reproduces 72% of bug recordings, significantly outperforming state-of-the-art baselines and ablation variants.

cs.SE↗

arXiv2026

The Role of Road Features and Vehicle Dynamics in Cost-Effective Autonomous Vehicles Safety Testing: Insights from Instance Space Analysis

Context: Simulation-based testing is a cost-efficient alternative to field testing for Autonomous Vehicles (AVs), but generating safety-critical test cases is challenging due to the vast search space. Prior work has studied static (road features) and dynamic (AV behavior) features of test scenarios separately, but their inter-dependencies are underexplored. Objective: In this paper, we describe an empirical to analyze how static and dynamic featuresof test scenarios, and their inter-dependencies, influence AV test scenario outcomes. Method: This study proposes an integrated approach using Instance Space Analysis (ISA) toevaluate both types of features, identify key influences on AV safety, and predict test outcomeswithout execution. Results: Our study identifies critical features affecting test outcomes (effective/ineffective, depending on whether it leads to a safety-critical condition). Results show that combining static and dynamic features improves prediction accuracy, confirmed by models trained on both feature types outperforming models trained with only one type of feature. Conclusion: The interplay of static and dynamic features enhances fault detection in AV testing. This research underscores the importance of integrating both types of features to create more effective testing frameworks for autonomous systems. Key contributions include: (1) a unified framework for AV safety assessment, (2) identification of influential features using ISA, and (3) efficient test outcome prediction for optimized regression testing.

cs.SE↗

arXiv2026

Unveiling Practical Shortcomings of Patch Overfitting Detection Techniques

Automated Program Repair (APR) can reduce the time developers spend debugging, allowing them to focus on other aspects of software development. Automatically generated bug patches are typically validated through software testing. However, this method can lead to patch overfitting, i.e., generating patches that pass the given tests but are still incorrect. Patch correctness assessment (also known as overfitting detection) techniques have been proposed to identify patches that overfit. However, prior work often assessed the effectiveness of these techniques in isolation and on datasets that do not reflect the distribution of correct-to-overfitting patches that would be generated by APR tools in typical use; thus, we still do not know their effectiveness in practice. This work presents the first comprehensive benchmarking study of several patch overfitting detection (POD) methods in a practical scenario. To this end, we curate datasets that reflect realistic assumptions (i.e., patches produced by tools run under the same experimental conditions). Next, we use these data to benchmark six state-of-the-art POD approaches -- spanning static analysis, dynamic testing, and learning-based approaches -- against two baselines based on random sampling (one from prior work and one proposed herein). Our results are striking: Simple random selection outperforms all POD tools for 71% to 96% of cases, depending on the POD tool. This suggests two main takeaways: (1) current POD tools offer limited practical benefit, highlighting the need for novel techniques; (2) any POD tool must be benchmarked on realistic data and against random sampling to prove its practical effectiveness. To this end, we encourage the APR community to continue improving POD techniques and to adopt our proposed methodology for practical benchmarking; we make our data and code available to facilitate such adoption.

cs.SE↗

arXiv2026

Empirical and Sustainability Aspects of Software Engineering Research in the Era of Large Language Models: A Reflection

Software Engineering (SE) research involving the use of Large Language Models (LLMs) has introduced several new challenges related to rigour in benchmarking, contamination, replicability, and sustainability. In this paper, we invite the research community to reflect on how these challenges are addressed in SE. Our results provide a structured overview of current LLM-based SE research at ICSE, highlighting both encouraging practices and persistent shortcomings. We conclude with recommendations to strengthen benchmarking rigour, improve replicability, and address the financial and environmental costs of LLM-based SE.

cs.SE↗

arXiv2025

A Semantic-based Optimization Approach for Repairing LLMs: Case Study on Code Generation

Language Models (LMs) are widely used in software engineering for code generation, but they may produce erroneous code. Rather than repairing outputs, a more thorough remedy is to address underlying model failures. LM repair offers a lightweight solution: it requires minimal data, lowers computational cost, and limits side effects. Unlike full retraining, LM repair focuses on applying tailored updates to targeted neurons, making it suitable for limited resources, high-performance demands, or strict safety requirements. In this paper, we propose Semantic Targeting for Analytical Repair (STAR), a novel semantic-based optimization method for repairing LLMs. STAR realizes the main operations of repairing LMs in an optimization process, including locating ``buggy neurons'', solving ``neuron patches'', and patching ``buggy neurons''. The neuron patches are computed with a solid semantic-based analytical formula, which directly bridges the changes to logits with the deltas of neurons, by steering latent representations. Compared to the prior work of LM repair (MINT) and standard optimization methods (SGD), STAR integrates their strengths while mitigating their limitations. By reformulating LM repair as an optimization process, STAR may solve multiple failures together, significantly improving the usefulness. Evaluated on coding tasks using popular code LMs, STAR demonstrates superior effectiveness compared with the state-of-the-art. Besides, STAR exhibits better efficiency. In terms of side effects, namely the balance between generalization and specificity, STAR outperforms prior work by a significant margin. Additionally, we conducted assessments on the overfitting risk of LM repair as well as the cumulative impact. Further, we analyzed the differences with pipeline-based methods and explained the reason why STAR is better and how it mitigated the common limitations of LM repair.

cs.SE↗

arXiv2025

Requirements-Driven Automated Software Testing: A Systematic Review

Automated software testing has significant potential to enhance efficiency and reliability within software development processes. However, its broader adoption faces considerable challenges, particularly concerning alignment between test generation methodologies and software requirements. REquirements-Driven Automated Software Testing (REDAST) addresses this gap by systematically leveraging requirements as the foundation for automated test artifact generation. This systematic literature review (SLR) critically examines the REDAST landscape, analyzing the current state of requirements input formats, transformation techniques, generated test artifacts, evaluation methods, and prevailing limitations. We conducted a thorough analysis of 156 relevant studies selected through a rigorous multi-stage filtering process from an initial collection of 27,333 papers sourced from six major research databases. Our findings highlight the predominance of functional requirements, model-based specifications, and natural language formats. Rule-based techniques are extensively utilized, while machine learning-based approaches remain relatively underexplored. Furthermore, most existing frameworks are sequential and dependent on singular intermediate representations, and while test cases, structured textual formats, and requirements coverage are common, full automation remains rare. We identify significant gaps related to automation completeness and dependency on input quality. This comprehensive synthesis provides a detailed overview of REDAST research and limitations, offering clear, evidence-based recommendations to guide future advancements in automated software testing.

cs.SE↗

arXiv2025

Multi-Modal Requirements Data-based Acceptance Criteria Generation using LLMs

Acceptance criteria (ACs) play a critical role in software development by clearly defining the conditions under which a software feature satisfies stakeholder expectations. However, manually creating accurate, comprehensive, and unambiguous acceptance criteria is challenging, particularly in user interface-intensive applications, due to the reliance on domain-specific knowledge and visual context that is not always captured by textual requirements alone. To address these challenges, we propose RAGcceptance M2RE, a novel approach that leverages Retrieval-Augmented Generation (RAG) to generate acceptance criteria from multi-modal requirements data, including both textual documentation and visual UI information. We systematically evaluated our approach in an industrial case study involving an education-focused software system used by approximately 100,000 users. The results indicate that integrating multi-modal information significantly enhances the relevance, correctness, and comprehensibility of the generated ACs. Moreover, practitioner evaluations confirm that our approach effectively reduces manual effort, captures nuanced stakeholder intent, and provides valuable criteria that domain experts may overlook, demonstrating practical utility and significant potential for industry adoption. This research underscores the potential of multi-modal RAG techniques in streamlining software validation processes and improving development efficiency. We also make our implementation and a dataset available.

cs.SE↗