A5035433961- Heat Transfer and Boiling Studies
- Heat Transfer and Optimization
- Topic Modeling
- Natural Language Processing Techniques
- Carbon and Quantum Dots Applications
- Nanocluster Synthesis and Applications
- Nanofluid Flow and Heat Transfer
- Advanced Photocatalysis Techniques
- Luminescence Properties of Advanced Materials
- Advanced Semiconductor Detectors and Materials
- Metal and Thin Film Mechanics
- Micro and Nano Robotics
- Advanced Sensor Technologies Research
- Innovation and Knowledge Management
- Metal Alloys Wear and Properties
- Solar Thermal and Photovoltaic Systems
- Text Readability and Simplification
- Solar-Powered Water Purification Methods
- Advanced X-ray Imaging Techniques
- Multimodal Machine Learning Applications
- Context-Aware Activity Recognition Systems
- Microwave Dielectric Ceramics Synthesis
- Software Engineering Research
- Advanced Optical Sensing Technologies
- Advanced Multi-Objective Optimization Algorithms
Yunnan University
2022-2025
University of Jinan
2015-2024
Hubei University of Medicine
2024
Taihe Hospital
2024
Tsinghua University
2022-2023
Qingdao University
2023
Center for Information Technology
2022
Open Group
2022
Hunan Agricultural University
2021
Fujitsu (Japan)
2016
Abstract As pre-trained language models (PLMs) have become the fundamental infrastructure for various NLP tasks and researchers readily enjoyed themselves in pretraining-finetuning paradigm, evidence from emerging research has continuously proven that larger tend to yield better performance. However, despite welcome outcome, process of fine-tuning large-scale PLMs brings prohibitive adaptation costs. In fact, fine- tuning all parameters a colossal model retaining separate instances different...
Prompt-learning has become a new paradigm in modern natural language processing, which directly adapts pre-trained models (PLMs) to $cloze$-style prediction, autoregressive modeling, or sequence generation, resulting promising performances on various tasks. However, no standard implementation framework of prompt-learning is proposed yet, and most existing codebases, often unregulated, only provide limited implementations for specific scenarios. Since there are many details such as templating...
Humans possess an extraordinary ability to create and utilize tools, allowing them overcome physical limitations explore new frontiers. With the advent of foundation models, AI systems have potential be equally adept in tool use as humans. This paradigm, i.e., learning with combines strengths specialized tools models achieve enhanced accuracy, efficiency, automation problem-solving. Despite its immense potential, there is still a lack comprehensive understanding key challenges,...
Green-emitting carbon quantum dots (G-CQDs) were prepared using tartaric acid and bran by one-pot solvothermal treatment had photoluminescence yields (PL QY) as high 46%. The morphology of the G-CQDs is characterized TEM, which shows average diameter approximately ∼4.85 nm. FT-IR spectra display presence -OH, C-N, N-H -COOH on surface G-CQDs. emission wavelength was ∼539 nm in case ∼450 excitation wavelength, corresponds to green fluorescence. Furthermore, used a fluorescent probe for...
The burgeoning interest in developing Large Language Models (LLMs) with up to trillion parameters has been met concerns regarding resource efficiency and practical expense, particularly given the immense cost of experimentation. This scenario underscores importance exploring potential Small (SLMs) as a resource-efficient alternative. In this context, we introduce MiniCPM, specifically 1.2B 2.4B non-embedding parameter variants, not only excel their respective categories but also demonstrate...
A novel and facile strategy was applied in the design fabrication of a micromotor-assisted dual-functional platform for sensitive detection rapid removal aniline water.
Breast cancer remains a leading cause of cancer-related mortality among women, with triple-positive breast (TPBC) being particularly aggressive subtype. GATA binding protein 3 (GATA3) plays crucial role in the luminal differentiation epithelium and T-cell differentiation. However, relationship between GATA3 immune infiltration TPBC unclear. This study collected analyzed data from The Cancer Genome Atlas (TCGA), METABRIC, GSE123845 databases. Univariate multivariate Cox regression analyses,...
While long-context inference is crucial for advancing large language model (LLM) applications, its prefill speed remains a significant bottleneck. Current approaches, including sequence parallelism strategies and compute reduction through approximate attention mechanisms, still fall short of delivering optimal efficiency. This hinders scaling the inputs to longer sequences processing queries in timely manner. To address this, we introduce APB, an efficient framework that leverages multi-host...
Detector research is rapidly advancing to meet the growing demands for long-wavelength infrared (LWIR) detectors in applications such as deep space exploration, medical imaging, meteorological detection, and thermal imaging. In this study, we propose a high-performance Type-II superlattices (T2SLs) avalanche photodiode (APD) with separate absorption multiplication (SAM) structure. Compared conventional LWIR detectors, device achieves over 100-fold increase responsivity at 77 K. This...
Despite the success, process of fine-tuning large-scale PLMs brings prohibitive adaptation costs. In fact, all parameters a colossal model and retaining separate instances for different tasks are practically infeasible. This necessitates new branch research focusing on parameter-efficient PLMs, dubbed as delta tuning in this paper. contrast with standard fine-tuning, only fine-tunes small portion while keeping rest untouched, largely reducing both computation storage Recent studies have...
Drafting-then-verifying decoding methods such as speculative are widely adopted training-free to accelerate the inference of large language models (LLMs). Instead employing an autoregressive process decode tokens sequentially, initially creates drafts with efficient small model. Then LLMs required conduct verification and correction in a non-autoregressive fashion minimize time overhead. Generating longer can lead even more significant speedups once verified, but also incurs substantial...
Abstract In this study, by adjusting sulfuric acid concentrations, tunable multicolour S/N‐carbon quantum dots (CQDs) were synthesized from waste foam as the raw material. The S/N‐CQDs presented blue, blue–green, green, green–yellow and yellow emission with an peak shifting 475 to 589 nm optimum excitation wavelengths of 385, 405, 440, 450, 500 nm, respectively. Using transmission electron microscopy, seen be spherical in morphology a size around 6–8 nm. Fourier transform infrared spectra...
PIN InGaAs short wavelength infrared (SWIR) focal plane array (FPA) detectors have attracted extensive attention due to their high detectivity, quantum efficiency, room temperature operation, low dark current, and good radiation resistance. Furthermore, FPA wide applications in many fields, such as aviation safety, biomedicine, camouflage recognition, night vision. Recently, research has been conducted on the extension of response spectrum from visible light (VIS) through InP substrate...
Abstract In this paper, two types of carbon quantum dot (CQDs) were prepared using biocompatible l ‐methionine as the source and urea nitrogen a one‐step hydrothermal treatment. By changing reaction solvents (deionized (DI) water dimethylformamide (DMF)), maximum emission resulting CQDs shifted from blue to red light. Specifically, wavelength moved 433 nm 625 following embedding new functional group (–CONH–) on surface CQDs. Photoluminescence yields with reached 64% 61%, respectively. The...
Shengding Hu, Ning Ding, Weilin Zhao, Xingtai Lv, Zhen Zhang, Zhiyuan Liu, Maosong Sun. Proceedings of the 61st Annual Meeting Association for Computational Linguistics (Volume 3: System Demonstrations). 2023.
The emergence of large language models (LLMs) relies heavily on distributed training strategies, among which pipeline parallelism plays a crucial role. As LLMs' sequence length extends to 32k or even 128k, the current parallel methods face severe bottlenecks, including high memory footprints and substantial bubbles, greatly hindering model scalability throughput. To enhance efficiency throughput, in this work, we introduce an efficient sequence-level one-forward-one-backward (1F1B)...