arXiv subjects

Mingfei Zhang

Publications and source records attributed to Mingfei Zhang.

4 recordsLinked to original sources

arXiv2026

OpenHarmony Bench: Evaluating LLMs and Coding Agents on OpenHarmony App Development

We present OPENHARMONY BENCH, an app-level coding benchmark for evaluating LLM-based coding agents on OpenHarmony ArkTS applications. Unlike function-level benchmarks, it evaluates complete app-level changes: each task requires an agent to modify a buildable ArkTS project so that a requested behavior works end to end, involving UI state, data persistence, build configuration, and platform APIs. The benchmark installs and drives the delivered application on a device to check whether the behavior is observable. It covers three input sources: natural-language feature requests (new-feature), structured scenario specifications (spec-driven), and bug descriptions (bug-fix). The benchmark contains 153 top-level tasks and 242 Feature points (F-points), where an F-point is one executable behavior check. The snapshot includes 32 new-feature tasks, 50 spec-driven tasks with 139 F-points, and 71 bug-fix tasks. The main leaderboard is scored over top-level tasks rather than independently weighted F-points. We describe the benchmark construction, statistics, and build-and-test evaluation pipeline, and evaluate DevEco Code with eight LLMs across three independent full-suite runs per configuration. Three findings emerge. First, newer generations complete more tasks than their predecessors within evaluated model-family pairs. Second, buildability is close to saturated while behavioral correctness is not: mean Final Build Success Rate is 94.77% to 100.00%, whereas mean Task Completion is 48.36% to 58.39%. Third, spec-driven tasks have the lowest Task Completion under all-checks task scoring, with no configuration exceeding 35%. The code, data, tasks, reference solutions, tests, evaluation scripts, and leaderboard are released through the official OPENHARMONY BENCH website at https://bench.matrix.openharmony.cn/.

cs.SE↗

arXiv2026

Beyond Explicit Refusals: Soft-Failure Attacks on Retrieval-Augmented Generation

Existing jamming attacks on Retrieval-Augmented Generation (RAG) systems typically induce explicit refusals or denial-of-service behaviors, which are conspicuous and easy to detect. In this work, we formalize a subtler availability threat, termed soft failure, which degrades system utility by inducing fluent and coherent yet non-informative responses rather than overt failures. We propose Deceptive Evolutionary Jamming Attack (DEJA), an automated black-box attack framework that generates adversarial documents to trigger such soft failures by exploiting safety-aligned behaviors of large language models. DEJA employs an evolutionary optimization process guided by a fine-grained Answer Utility Score (AUS), computed via an LLM-based evaluator, to systematically degrade the certainty of answers while maintaining high retrieval success. Extensive experiments across multiple RAG configurations and benchmark datasets show that DEJA consistently drives responses toward low-utility soft failures, achieving SASR above 79\% while keeping hard-failure rates below 15\%, significantly outperforming prior attacks. The resulting adversarial documents exhibit high stealth, evading perplexity-based detection and resisting query paraphrasing, and transfer across model families to proprietary systems without retargeting.

cs.CR↗

arXiv2022

Mechanism of Local Lattice Distortion Effects on Vacancy Migration Barriers in FCC Alloys

Accurate prediction of vacancy migration energy barriers, $ΔE_a$, in multi-component alloys is extremely challenging yet critical for the development of diffusional transformation kinetics needed to model alloy behavior in many technological applications. Here, results from $ΔE_a$ and the energy driving force $ΔE$ of many (>1000) vacancy migration events calculated using density functional theory and nudged elastic band method show large changes (~1eV) of $ΔE_a$ in different local chemical environments of the model face-centered cubic Al-Mg-Zn alloys. Due to local lattice distortion effects induced by solute atoms (such as Mg) with different sizes than the matrix element (Al), the changes of $ΔE_a$ for one type of migrating atoms originate primarily from fluctuations of $Δe_a\equiv ΔE_a - \frac{1}{2}ΔE$. To understand the fluctuations, a quartic function is shown to accurately describe the energy landscape of the minimum energy path (MEP) for each vacancy migration event. Analyses of the quartic function show that $Δe_a$ can be approximated with $Δe_a \approx αk_fD^2$, where $α\sim 0.022$ is a constant of all types of migrating atoms. Here $D$ is the distance of a migrating atom between two adjacent equilibrium positions and $k_f$ is the average vibration spring constant of this atom at these two equilibrium positions. $k_f$ and $D$ quantitatively describe the lattice distortion effects on the curvatures and locations of the MEP at its initial and final states in different local chemical environments. We also used the local lattice occupations as inputs to train surrogate models to predict coefficients of the quartic function, which accurately and efficiently output both $ΔE_a$ and $ΔE$ as the necessary inputs for the mesoscale studies of diffusional transformation in Al-Mg-Zn alloys.

cond-mat.mtrl-sci↗

arXiv2019

Predicting densities and elastic moduli of SiO2-based glasses by machine learning

Chemical design of SiO2-based glasses with high elastic moduli and low weight is of great interest. However, it is difficult to find a universal expression to predict the elastic moduli according to the glass composition before synthesis since the elastic moduli are a complex function of interatomic bonds and their ordering at different length scales. Here we show that the densities and elastic moduli of SiO2-based glasses can be efficiently predicted by machine learning (ML) techniques across a complex compositional space with multiple (>10) types of additive oxides besides SiO2. Our machine learning approach relies on a training set generated by high-throughput molecular dynamic (MD) simulations, a set of elaborately constructed descriptors that bridges the empirical statistical modeling with the fundamental physics of interatomic bonding, and a statistical learning/predicting model developed by implementing least absolute shrinkage and selection operator with a gradient boost machine (GBM-LASSO). The predictions of the ML model are comprehensively compared and validated with a large amount of both simulation and experimental data. By just training with a dataset only composed of binary and ternary glass samples, our model shows very promising capabilities to predict the density and elastic moduli for k-nary SiO2-based glasses beyond the training set. As an example of its potential applications, our GBM-LASSO model was used to perform a rapid and low-cost screening of many (~105) compositions of a multicomponent glass system to construct a compositional-property database that allows for a fruitful overview on the glass density and elastic properties.

cond-mat.mtrl-sci↗