A5020225061- Topic Modeling
- Photonic and Optical Devices
- Natural Language Processing Techniques
- Advanced Fiber Laser Technologies
- Aluminum Alloys Composites Properties
- Advanced Graph Neural Networks
- Microstructure and mechanical properties
- Aluminum Alloy Microstructure Properties
- Speech and dialogue systems
- Advanced Photonic Communication Systems
- Multimodal Machine Learning Applications
- Distributed systems and fault tolerance
- Advanced Data Storage Technologies
- Security and Verification in Computing
- Caching and Content Delivery
- Physical Unclonable Functions (PUFs) and Hardware Security
- Neural Networks and Applications
- Acoustic Wave Resonator Technologies
- Radiation Effects in Electronics
- Neural Networks and Reservoir Computing
- Seismology and Earthquake Studies
- Semantic Web and Ontologies
- Phytochemistry and Bioactive Compounds
- Dielectric materials and actuators
- MXene and MAX Phase Materials
Harbin Medical University
2025
Shanghai Jiao Tong University
2016-2024
Tencent (China)
2024
State Grid Corporation of China (China)
2024
The University of Sydney
2024
Tsinghua University
2018-2022
Harbin University of Science and Technology
2020
Jilin University
2019
University Town of Shenzhen
2018
Zhi Chen, Lu Yanbin Zhao, Ruisheng Cao, Zihan Xu, Su Zhu, Kai Yu. Proceedings of the 2021 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2021.
In situ TiC/Al-Cu3.7-Mg1.3 nanocomposites containing 5–20 vol. % nano-TiC with nacre-like structures were prepared via combustion synthesis and hot pressing assisted by extrusion. Dry sliding friction wear tests of the in Al-Cu3.7-Mg1.3 alloy was carried out a pin-on-disc apparatus under various applied loads at different velocities room temperature. The showed much better performance than alloy. SEM EDS analyses conducted to identify mechanisms. enhanced resistance attributed formation...
Integrated electro-optic modulators are key components in photonic integrated circuits. Silicon technology is considered to be promising for large-scale and low-cost integration. However, silicon does not exhibit any Pockels effect, the modulator based on free-carrier dispersion suffers from challenges such as high-power consumption, limited bandwidth, large optical propagation loss. Here, a new, best of our knowledge, hybrid lithium tantalite-silicon platform proposed effect. Benefiting...
Deep question generation (DQG) aims to generate complex questions through reasoning over multiple documents. The task is challenging and underexplored. Existing methods mainly focus on enhancing document representations, with little attention paid the answer information, which may result in generated not matching type being answerirrelevant. In this paper, we propose an Answer-driven Question Generation (ADDQG) model based encoder-decoder framework. makes better use of target as a guidance...
Nowadays many practical systems adopt erasure codes to ensure reliability and reduce storage overhead. However, also bring in low recovery performance. The network links practice, such as peer-to-peer cross-data network, always have nonuniform bandwidth because of various reasons. To time, we propose Parallel Pipeline Tree (PPT) Cross-Tree (PPCT) speed up single-node multiple-node non-uniform traffic environment, respectively. By utilizing gap among links, PPT constructs a tree path based on...
Abstract On-chip optical communications are growingly aiming at multimode operation together with mode-division multiplexing to further increase the transmission capacity. Optical switches, which capable of signals switching nodes, play a key role in networks. We demonstrate 2 × electro-optic Mach–Zehnder interferometer-based mode- and polarization-selective switch fabricated by standard complementary metal–oxide–semiconductor process. An electro optic tuner based on PN-doped junction one...
Abstract Power consumption of photonic integrated circuits becomes a critical consideration. A new platform is proposed for ultralow‐power tuning in silicon photonics via piezo‐optomechanical coupling using hafnium‐oxide actuators. As an example the potential platform, tunable silicon‐hafnium‐oxide hybrid microring, where film acts as active optical and piezoelectric layer, demonstrated. The microring capable linear bidirectional with wavelength efficiency 8.4 pm V −1 power 0.12 nW ....
In dialogue systems, a state tracker aims to accurately find compact representation of the current status, based on entire history. While previous approaches often define states as combination separate triples ({\em domain-slot-value}), in this paper, we employ structured and cast tracking sequence generation problem. Based new formulation, propose {\bf C}oa{\bf R}s{\bf E}-to-fine DI}alogue T}racking ({\bf CREDIT}) approach. Taking advantage representation, which is marked language sequence,...
Abstract Metalens has emerged as an important optical block in free-space systems, which shows excellent performance. Even the metalens based on gradient index (GRIN) profiles can be implemented for on-chip beam focusing behavior. However, most previous schemes, GRIN metalenses only achieve behavior one dimension, limits their applications low-loss waveguide interconnecting or fiber-to-chip coupling. In this paper, half Maxwell’s fisheye lens with subwavelength features, integrated silicon...
Aluminum nitride has advantages ranging from a large transparency window to its high thermal and chemical resistance, piezoelectric effect, electro-optic property, compatibility with the complementary metal-oxide-semiconductor fabrication process. We propose hybrid aluminum silicon platform for integrated photonics. Hybrid nitride-silicon basic photonic devices, including multimode interferometer, Mach-Zehnder micro-ring resonator, are designed fabricated. The measured extinction ratio is >...
Abstract Integrated tunable optical filters are essential components in photonic signal processors, telecom systems, sensors, and quantum devices. Two of the most important features a filter its dimensions power consumption. Herein, design experimental validation an on‐chip composed 1D topological crystal cavity based on hybrid integrated lithium tantalite‐silicon platform is presented. The strong confinement boundary state allows fabrication with ultra‐compact size 1.14 × 75 µm 2...
Relation detection is a key step in Knowledge Base Question Answering (KBQA), but far from solved due to the significant differences between questions and relations. Previous studies usually treat relation as text matching task, mainly focus on reducing error with better representations of KB However, understanding also important since they are generally more varied. And pair representation requires improvement because relations not always counterparts questions. In this paper, we propose...
Semantic matching is a basic problem in natural language processing, but it far from solved because of the differences between pairs for matching. In question answering (QA), answer selection (AS) popular semantic task, usually reformulated as paraphrase identification (PI) problem. However, QA different PI and are not synonymous sentences strictly comparable. this work, novel knowledge cross-pair pattern guided system (KCG) proposed, which considers both conditions QA. We apply explicit...
Retrieval augmentation has become an effective solution to empower large language models (LLMs) with external and verified knowledge sources from the database, which overcomes limitations hallucinations of LLMs in handling up-to-date domain-specific information. However, existing embedding for text retrieval usually have three non-negligible limitations. First, number diversity samples a batch are too restricted supervise modeling textual nuances at scale. Second, high proportional noise...
Knowledge distillation (KD) has been widely adopted to compress large language models (LLMs). Existing KD methods investigate various divergence measures including the Kullback-Leibler (KL), reverse (RKL), and Jensen-Shannon (JS) divergences. However, due limitations inherent in their assumptions definitions, these fail deliver effective supervision when few distribution overlap exists between teacher student. In this paper, we show that aforementioned KL, RKL, JS divergences respectively...
Knowledge distillation (KD) has been widely adopted to compress large language models (LLMs). Existing KD methods investigate various divergence measures including the Kullback-Leibler (KL), reverse KL (RKL), and Jensen-Shannon (JS) divergences. However, due limitations inherent in their assumptions definitions, these fail deliver effective supervision when a distribution overlap exists between teacher student. In this article, we show that aforementioned KL, RKL, JS divergences,...