Multi-Modal Anomaly Detection: A Survey
本文从假设驱动的角度综述了多模态异常检测问题，通过正常性假设和异常性假设两种方法解决跨数据源的罕见异常事件检测。
0 citationsRead paper
Selected work
Representative Papers
Scaling Unmodified Multithreaded Applications with Elastic CXL-based Distributed Shared Memory
This work addresses the challenge of transparently scaling multithreaded applications on existing CXL-based distributed shared memory (DSM) systems, which are hindered by the need for manual code modifications, static data placement policies, and sub-microsecond page fault overheads. To overcome these limitations, the authors propose a co-designed operating system and runtime that establish a global, unified address space, enabling transparent application scaling without source-code changes. Key innovations include a dynamic, latency-driven data placement strategy and an elastic page management mechanism that supports online page merging and splitting while exploiting spatial locality. Experimental evaluation across 15 configurations and diverse workloads demonstrates performance improvements of 1.5–2.2× over a pure CXL baseline and 1.1–2.2× over state-of-the-art hybrid DSM systems, achieving near-linear scalability.
0 citationsRead paper
A Generalizable Framework for Building Executable Domain-Specific LLMs under Data Scarcity: Demonstration on Semiconductor TCAD Simulation
This work addresses the challenge of developing domain-specific large language models (LLMs) capable of generating both fluent natural language and executable code in data-scarce scientific and engineering domains. The authors propose a pattern-first alignment framework that injects domain knowledge through large-scale synthetic question-answering pairs and innovatively integrates intermediate representation (IR)-driven direct preference optimization (DPO) with controllable retrieval-augmented generation (RAG). This approach jointly enhances instruction following and code executability under low-resource conditions. Evaluated on the TCAD task, the method achieves 85.6% semantic accuracy and 80.0% syntactic pass rate, significantly outperforming GPT-4o. Its transferability is further validated on the Elmer solver, establishing a reproducible and generalizable paradigm for building domain-specific LLMs.
0 citationsRead paper
Dynamic Facial Expressions Analysis Based Parkinson's Disease Auxiliary Diagnosis
Early Parkinson’s disease (PD) lacks non-invasive, convenient screening tools—particularly for objective quantification of hypomimia, a hallmark facial bradykinesia. To address this, we propose a PD辅助 diagnosis framework based on dynamic facial expression analysis. Our method is the first to adapt the CLIP multimodal architecture for temporal modeling of facial expression intensity, leveraging vision–text semantic priors to enhance PD-specific feature representation. We further integrate an LSTM-based temporal classifier with a novel dynamic expression intensity normalization strategy to improve robustness across subjects and sessions. Evaluated on a clinical dataset, our approach achieves 93.1% diagnostic accuracy—significantly outperforming existing ex vivo biomarker assays. It enables remote initial screening and longitudinal home-based monitoring, establishing a new paradigm for contactless, intelligent PD screening.
0 citationsRead paper
Graph Convolutional Long Short-Term Memory Attention Network for Post-Stroke Compensatory Movement Detection Based on Skeleton Data
To address the challenge of accurately identifying compensatory movements during stroke rehabilitation, this paper proposes a Graph Convolutional Network–Long Short-Term Memory–Attention (GCN-LSTM-ATT) fusion model. Leveraging skeletal sequence data captured by Kinect, the model employs GCN to encode spatial topological relationships among joints, LSTM to capture temporal dynamics, and an attention mechanism to enhance discriminative capability for critical motion segments. Ablation studies confirm the effectiveness of each component. Evaluated on a real-world stroke rehabilitation dataset, the model achieves an accuracy of 0.8580—significantly outperforming conventional methods including SVM, KNN, and Random Forest. This work establishes a novel paradigm for automated, fine-grained identification of compensatory motions, thereby enabling data-driven personalization and optimization of rehabilitation strategies.
0 citationsRead paper
Recent publications
Latest Papers
Multi-Modal Anomaly Detection: A Survey
本文从假设驱动的角度综述了多模态异常检测问题，通过正常性假设和异常性假设两种方法解决跨数据源的罕见异常事件检测。
0 citationsRead paper
Scaling Unmodified Multithreaded Applications with Elastic CXL-based Distributed Shared Memory
This work addresses the challenge of transparently scaling multithreaded applications on existing CXL-based distributed shared memory (DSM) systems, which are hindered by the need for manual code modifications, static data placement policies, and sub-microsecond page fault overheads. To overcome these limitations, the authors propose a co-designed operating system and runtime that establish a global, unified address space, enabling transparent application scaling without source-code changes. Key innovations include a dynamic, latency-driven data placement strategy and an elastic page management mechanism that supports online page merging and splitting while exploiting spatial locality. Experimental evaluation across 15 configurations and diverse workloads demonstrates performance improvements of 1.5–2.2× over a pure CXL baseline and 1.1–2.2× over state-of-the-art hybrid DSM systems, achieving near-linear scalability.
0 citationsRead paper
A Generalizable Framework for Building Executable Domain-Specific LLMs under Data Scarcity: Demonstration on Semiconductor TCAD Simulation
This work addresses the challenge of developing domain-specific large language models (LLMs) capable of generating both fluent natural language and executable code in data-scarce scientific and engineering domains. The authors propose a pattern-first alignment framework that injects domain knowledge through large-scale synthetic question-answering pairs and innovatively integrates intermediate representation (IR)-driven direct preference optimization (DPO) with controllable retrieval-augmented generation (RAG). This approach jointly enhances instruction following and code executability under low-resource conditions. Evaluated on the TCAD task, the method achieves 85.6% semantic accuracy and 80.0% syntactic pass rate, significantly outperforming GPT-4o. Its transferability is further validated on the Elmer solver, establishing a reproducible and generalizable paradigm for building domain-specific LLMs.
0 citationsRead paper
Dynamic Facial Expressions Analysis Based Parkinson's Disease Auxiliary Diagnosis
Early Parkinson’s disease (PD) lacks non-invasive, convenient screening tools—particularly for objective quantification of hypomimia, a hallmark facial bradykinesia. To address this, we propose a PD辅助 diagnosis framework based on dynamic facial expression analysis. Our method is the first to adapt the CLIP multimodal architecture for temporal modeling of facial expression intensity, leveraging vision–text semantic priors to enhance PD-specific feature representation. We further integrate an LSTM-based temporal classifier with a novel dynamic expression intensity normalization strategy to improve robustness across subjects and sessions. Evaluated on a clinical dataset, our approach achieves 93.1% diagnostic accuracy—significantly outperforming existing ex vivo biomarker assays. It enables remote initial screening and longitudinal home-based monitoring, establishing a new paradigm for contactless, intelligent PD screening.
0 citationsRead paper
Graph Convolutional Long Short-Term Memory Attention Network for Post-Stroke Compensatory Movement Detection Based on Skeleton Data
To address the challenge of accurately identifying compensatory movements during stroke rehabilitation, this paper proposes a Graph Convolutional Network–Long Short-Term Memory–Attention (GCN-LSTM-ATT) fusion model. Leveraging skeletal sequence data captured by Kinect, the model employs GCN to encode spatial topological relationships among joints, LSTM to capture temporal dynamics, and an attention mechanism to enhance discriminative capability for critical motion segments. Ablation studies confirm the effectiveness of each component. Evaluated on a real-world stroke rehabilitation dataset, the model achieves an accuracy of 0.8580—significantly outperforming conventional methods including SVM, KNN, and Random Forest. This work establishes a novel paradigm for automated, fine-grained identification of compensatory motions, thereby enabling data-driven personalization and optimization of rehabilitation strategies.
0 citationsRead paper