E-CONAN (Entailment, CONtradition And Neutral) Benchmarks: Arabic Textual Entailment and Natural Inference Datasets
为解决阿拉伯语在自然语言推理资源有限的问题，本文通过多种来源构建E-CONAN基准数据集，并用其评估了多语言预训练模型和特定于阿拉伯语的模型。
0 citationsRead paper
Selected work
Representative Papers
Context-Aware Web Attack Detection in Open-Source SIEM Systems via MITRE ATT&CK-Enriched Behavioral Profiling
This study addresses the limitations of traditional rule-based SIEM systems in detecting multi-step web attacks due to insufficient contextual information about source host behavior. To overcome this, the authors propose Smart-SIEM, an enhanced framework integrated into the Wazuh platform that constructs a novel source IP behavioral context vector by fusing HTTP response status codes, rule-trigger frequencies, and MITRE ATT&CK technique occurrences. A two-stage cascaded model combining LightGBM and XGBoost is designed for attack detection and fine-grained classification, complemented by an adaptive retraining mechanism to mitigate concept drift. Experimental results demonstrate a binary classification F1-score of 0.967 and a six-class attack classification F1-score of 0.914, with perfect (100%) detection of brute-force attacks and 98.3% recall for authentication bypass attempts. When confronted with previously unseen attacks, the system’s F1-score recovers from 0.465 to 0.814 after adaptive retraining.
0 citationsRead paper
Improving Cross-Patient Generalization in Parkinson's Disease Detection through Chunk-Based Analysis of Hand-Drawn Patterns
To address the limited cross-subject generalizability of Parkinson’s disease (PD) detection models, this paper proposes a multi-stage robust recognition method based on hand-drawn graphics. The approach innovatively introduces a 2×2 image tiling strategy, integrates graphic-type prior classification, and hierarchically extracts both local and global features. An ensemble decision mechanism fuses discriminative information from multiple sources. This design significantly mitigates performance degradation caused by inter-subject variability: on the NewHandPD dataset, the method achieves 97.08% accuracy for seen subjects and maintains 94.91% for unseen subjects—a mere 2.17-percentage-point gap—outperforming existing state-of-the-art methods. The core contribution lies in the synergistic integration of structured tiling, task-guided feature disentanglement, and ensemble learning, thereby enhancing both cross-subject generalizability and clinical applicability.
0 citationsRead paper
Survey of NLU Benchmarks Diagnosing Linguistic Phenomena: Why not Standardize Diagnostics Benchmarks?
Current NLU diagnostic benchmarks lack standardization: inconsistent naming conventions, no consensus on linguistic phenomena coverage, and absence of ISO-style evaluation protocols—hindering cross-benchmark comparability and fine-grained error analysis. This study addresses these gaps via a systematic literature review and cross-benchmark comparative analysis, conducting the first linguistic-phenomenon-level mapping and structured taxonomy of English, Arabic, and multilingual diagnostic datasets. Its core contributions are threefold: (1) identifying two critical deficiencies—naming inconsistency and incomplete phenomenon coverage; (2) proposing a globally scalable, hierarchical classification system for linguistic phenomena in NLU diagnostics; and (3) pioneering a standardized evaluation framework for NLU diagnostics, modeled after industrial standards, to enable theoretically grounded and practically actionable fine-grained capability assessment.
0 citationsRead paper
Parallel Paradigms in Modern HPC: A Comparative Analysis of MPI, OpenMP, and CUDA
Selecting appropriate parallel programming models for heterogeneous HPC architectures remains challenging due to divergent hardware characteristics and software trade-offs. Method: This paper conducts the first multi-dimensional quantitative comparison of MPI, OpenMP, and CUDA—evaluating architectural adaptability, scalability bottlenecks, development complexity, and domain suitability—and proposes a hybrid programming model selection framework tailored to heterogeneity. The framework integrates communication modeling, memory contention analysis, and GPU kernel optimization for empirical validation. Contribution/Results: Experiments show MPI achieves >92% strong scaling efficiency in distributed, communication-intensive workloads; OpenMP delivers 3.8× speedup on shared-memory loop-parallel tasks; CUDA attains up to 12.5× acceleration on data-parallel kernels; and hybrid strategies yield an average 27% improvement in end-to-end performance. The study provides both theoretical foundations and practical guidelines for optimizing and co-designing programming models in heterogeneous HPC environments.
0 citationsRead paper
Recent publications
Latest Papers
E-CONAN (Entailment, CONtradition And Neutral) Benchmarks: Arabic Textual Entailment and Natural Inference Datasets
为解决阿拉伯语在自然语言推理资源有限的问题，本文通过多种来源构建E-CONAN基准数据集，并用其评估了多语言预训练模型和特定于阿拉伯语的模型。
0 citationsRead paper
Context-Aware Web Attack Detection in Open-Source SIEM Systems via MITRE ATT&CK-Enriched Behavioral Profiling
This study addresses the limitations of traditional rule-based SIEM systems in detecting multi-step web attacks due to insufficient contextual information about source host behavior. To overcome this, the authors propose Smart-SIEM, an enhanced framework integrated into the Wazuh platform that constructs a novel source IP behavioral context vector by fusing HTTP response status codes, rule-trigger frequencies, and MITRE ATT&CK technique occurrences. A two-stage cascaded model combining LightGBM and XGBoost is designed for attack detection and fine-grained classification, complemented by an adaptive retraining mechanism to mitigate concept drift. Experimental results demonstrate a binary classification F1-score of 0.967 and a six-class attack classification F1-score of 0.914, with perfect (100%) detection of brute-force attacks and 98.3% recall for authentication bypass attempts. When confronted with previously unseen attacks, the system’s F1-score recovers from 0.465 to 0.814 after adaptive retraining.
0 citationsRead paper
Improving Cross-Patient Generalization in Parkinson's Disease Detection through Chunk-Based Analysis of Hand-Drawn Patterns
To address the limited cross-subject generalizability of Parkinson’s disease (PD) detection models, this paper proposes a multi-stage robust recognition method based on hand-drawn graphics. The approach innovatively introduces a 2×2 image tiling strategy, integrates graphic-type prior classification, and hierarchically extracts both local and global features. An ensemble decision mechanism fuses discriminative information from multiple sources. This design significantly mitigates performance degradation caused by inter-subject variability: on the NewHandPD dataset, the method achieves 97.08% accuracy for seen subjects and maintains 94.91% for unseen subjects—a mere 2.17-percentage-point gap—outperforming existing state-of-the-art methods. The core contribution lies in the synergistic integration of structured tiling, task-guided feature disentanglement, and ensemble learning, thereby enhancing both cross-subject generalizability and clinical applicability.
0 citationsRead paper
Survey of NLU Benchmarks Diagnosing Linguistic Phenomena: Why not Standardize Diagnostics Benchmarks?
Current NLU diagnostic benchmarks lack standardization: inconsistent naming conventions, no consensus on linguistic phenomena coverage, and absence of ISO-style evaluation protocols—hindering cross-benchmark comparability and fine-grained error analysis. This study addresses these gaps via a systematic literature review and cross-benchmark comparative analysis, conducting the first linguistic-phenomenon-level mapping and structured taxonomy of English, Arabic, and multilingual diagnostic datasets. Its core contributions are threefold: (1) identifying two critical deficiencies—naming inconsistency and incomplete phenomenon coverage; (2) proposing a globally scalable, hierarchical classification system for linguistic phenomena in NLU diagnostics; and (3) pioneering a standardized evaluation framework for NLU diagnostics, modeled after industrial standards, to enable theoretically grounded and practically actionable fine-grained capability assessment.
0 citationsRead paper
Parallel Paradigms in Modern HPC: A Comparative Analysis of MPI, OpenMP, and CUDA
Selecting appropriate parallel programming models for heterogeneous HPC architectures remains challenging due to divergent hardware characteristics and software trade-offs. Method: This paper conducts the first multi-dimensional quantitative comparison of MPI, OpenMP, and CUDA—evaluating architectural adaptability, scalability bottlenecks, development complexity, and domain suitability—and proposes a hybrid programming model selection framework tailored to heterogeneity. The framework integrates communication modeling, memory contention analysis, and GPU kernel optimization for empirical validation. Contribution/Results: Experiments show MPI achieves >92% strong scaling efficiency in distributed, communication-intensive workloads; OpenMP delivers 3.8× speedup on shared-memory loop-parallel tasks; CUDA attains up to 12.5× acceleration on data-parallel kernels; and hybrid strategies yield an average 27% improvement in end-to-end performance. The study provides both theoretical foundations and practical guidelines for optimizing and co-designing programming models in heterogeneous HPC environments.
0 citationsRead paper