A5101188063- Topic Modeling
- Natural Language Processing Techniques
- Photonic and Optical Devices
- Multimodal Machine Learning Applications
- Speech and dialogue systems
- Software Engineering Research
- Advanced Photonic Communication Systems
- Domain Adaptation and Few-Shot Learning
- Software Reliability and Analysis Research
- Magnesium Alloys: Properties and Applications
- Optical Network Technologies
- Advanced Text Analysis Techniques
- Hydrogen Storage and Materials
- Machine Learning and Data Classification
- Photonic Crystals and Applications
- Wastewater Treatment and Nitrogen Removal
- Water Treatment and Disinfection
- Text Readability and Simplification
- Antenna Design and Optimization
- Software Testing and Debugging Techniques
- Advanced Graph Neural Networks
- Circular RNAs in diseases
- Thermal Radiation and Cooling Technologies
- Intelligent Tutoring Systems and Adaptive Learning
- Web Data Mining and Analysis
University of Michigan–Dearborn
2022-2025
Intel (United States)
2025
Michigan United
2025
Alibaba Group (United States)
2023-2024
University of Science and Technology Beijing
2023-2024
Wuhan University
2020-2024
Nanjing University
2023-2024
Institute of Geology, China Earthquake Administration
2024
ETH Zurich
2024
Harbin Institute of Technology
2023-2024
Large language models (LLMs) often contain misleading content, emphasizing the need to align them with human values ensure secure AI systems. Reinforcement learning from feedback (RLHF) has been employed achieve this alignment. However, it encompasses two main drawbacks: (1) RLHF exhibits complexity, instability, and sensitivity hyperparameters in contrast SFT. (2) Despite massive trial-and-error, multiple sampling is reduced pair-wise contrast, thus lacking contrasts a macro perspective. In...
Circular RNAs (circRNAs) are covalently closed RNA structures that play multiple roles in tumorigenesis and progression. Compared with exon‒intron circRNAs, the biological functions implications of intergenic circRNAs human cancer still poorly understood. Here, we performed circRNA microarray analysis identified an circRNA,
Recent research has demonstrated that Large Language Models (LLMs) can enhance their capabilities by utilizing external tools. However, three pivotal questions remain unanswered: (1) How effective are current LLMs in tools? (2) we LLMs' ability to utilize (3) What obstacles need be overcome leverage To address these questions, introduce API-Bank, a groundbreaking benchmark, specifically designed for tool-augmented LLMs. For the first question, develop runnable evaluation system consisting of...
Large language models (LLMs) have revolutionized the field of artificial intelligence, enabling natural processing tasks that were previously thought to be exclusive humans. In this work, we introduce Qwen, first installment our large model series. Qwen is a comprehensive series encompasses distinct with varying parameter counts. It includes base pretrained models, and Qwen-Chat, chat finetuned human alignment techniques. The consistently demonstrate superior performance across multitude...
Optical phased arrays (OPAs) are renowned for their exceptional ability to manipulate and direct light beams, positioning them as an ideal choice solid-state detection ranging (LiDAR) technologies. This review provides a comprehensive examination of the current research landscape photonic integrated circuit (PIC)-based OPAs. It begins by addressing critical design elements at component level necessary optimal functionality. then delves into phase calibration techniques overarching...
Wastewater treatment is an important source of greenhouse gases (GHGs). Yet large uncertainties remain in the quantification GHG emissions from municipal wastewater plants (MWWTPs) China. A high-resolution and technology-specific emission inventory still lacking to support mitigation strategies MWWTPs. Here we develop a plant-level technology-based MWWTP for China covering 8703 19 technology categories by compiling harmonizing most up-to-date facility-level databases. China's methane (CH
Critiques are important for enhancing the performance of Large Language Models (LLMs), enabling both self-improvement and constructive feedback others by identifying flaws suggesting improvements. However, evaluating critique capabilities LLMs presents a significant challenge due to open-ended nature task. In this work, we introduce new benchmark designed assess LLMs. Unlike existing benchmarks, which typically function in an open-loop fashion, our approach employs closed-loop methodology...
Various approaches are proposed to help under-resourced security researchers detect and analyze software vulnerabilities. It is still incredibly time-consuming labor-intensive for fix such reported vulnerabilities due the increasing size complexity of modern systems. The time lag between reporting fixing a vulnerability causes systems suffer from significant exposure possible attacks. Very recently, some techniques propose apply pretrained models have proved their success in improving repair...
Automated program repair (APR) aims to fix software bugs without manual debugging efforts and plays a crucial role in development maintenance. Template-based APR has been widely investigated shown promising results. However, it is challenging for template-based select the appropriate donor code, which an important ingredient generating candidate patches. Inappropriate code may cause plausible but incorrect patch generation even with correct patterns, limiting performance. In this paper, we...
Large language models (LLMs) often contain misleading content, emphasizing the need to align them with human values ensure secure AI systems. Reinforcement learning from feedback (RLHF) has been employed achieve this alignment. However, it encompasses two main drawbacks: (1) RLHF exhibits complexity, instability, and sensitivity hyperparameters in contrast SFT. (2) Despite massive trial-and-error, multiple sampling is reduced pair-wise contrast, thus lacking contrasts a macro perspective. In...
Measuring the quality of responses generated by LLMs is a challenging task, particularly when it comes to evaluating whether response aligned with human preference. A novel approach involves using LLM itself make evaluation and stabilizing results through multiple independent evaluations, similar single-layer narrow network. This network consists fixed number neurons, each neuron being same LLM. In this paper, we draw upon extensive research on deep neural networks explore deeper wider can...
In this work, the combined addition of strontium/indium (Sr/In) to magnesium anode for Mg-Air Cells is investigated improve discharge performance by modifying anode/electrolyte interface. Indium exists as solid solution atoms in α-Mg matrix without its second-phase generation, and at same time facilitates grain refinement, dendritic segregation Mg17Sr2-phases precipitation. During operation, Sr modifies film composition via compounds promoted redeposition substrate/film interface; their...
Recent research has demonstrated that Large Language Models (LLMs) can enhance their capabilities by utilizing external tools. However, three pivotal questions remain unanswered: (1) How effective are current LLMs in tools? (2) we LLMs' ability to utilize (3) What obstacles need be overcome leverage To address these questions, introduce API-Bank, a groundbreaking benchmark, specifically designed for tool-augmented LLMs. For the first question, develop runnable evaluation system consisting of...
Large language models (LLMs) demonstrate remarkable ability to comprehend, reason, and generate following nature instructions. However, the development of LLMs has been primarily focused on high-resource languages, such as English, thereby limiting their applicability research in other languages. Consequently, we present PolyLM, a multilingual LLM trained 640 billion (B) tokens, avaliable two model sizes: 1.7B 13B. To enhance its capabilities, 1) integrate bilingual data into training data;...
The Rocahepevirus ratti hepatitis E virus genotype C1 (HEV-C1) has been documented to infect humans. However, the understanding of HEV-C1 remains constrained. This study aims determine prevalence and genomic characteristics in small animals Yunnan province southwestern China. A total 444 liver tissues were collected from covering orders Rodentia, Soricomorpha, Scandentia Erinaceomorpha three regions Yunnan. Then Paslahepevirus balayani examined using RT-qPCR. detection rate was 12.95 %...
Open Information Extraction (OpenIE) facilitates domain-independent discovery of relational facts from large corpora. The technique well suits many open-world natural language understanding scenarios, such as automatic knowledge base construction, open-domain question answering, and explicit reasoning. Thanks to the rapid development in deep learning technologies, numerous neural OpenIE architectures have been proposed achieve considerable performance improvement. In this survey, we provide...
This paper presents an experimental study on the emergence of central recirculation zone (CRZ) in a basic V-shaped premixed swirling flame, using simultaneous measurement particle image velocimetry and CH* chemiluminescence. Unlike most previous works that explored CRZ formation controlled by vortex breakdown (VB), present experiment maintains constant swirl intensity to rule out influence VB. We find characteristics its are strongly affected strength inner shear layer (ISL) surrounding CRZ....
Unit testing aims to validate the correctness of software system units and has become an essential practice in development maintenance. However, it is incredibly time-consuming labor-intensive for experts write unit test cases manually, including inputs ( i.e., prefixes) oracles assertions). Very recently, some techniques have been proposed apply Large Language Models (LLMs) generate assertions proven potential reducing manual efforts. there no systematic comparison effectiveness these LLMs,...
Bacterial skin infections cause a variety of common diseases that require drugs are safer than antibiotics and have fewer side effects. However, for evaluating disease drugs, human tissue in vitro constructed traditionally on Transwell has inefficient screening ability because its fragile barrier function. With mechanical forces dynamic flow, the organ-on-a-chip system became an innovative, automatic, modular way to construct pathological models analyze effective pharmaceutical ingredients...