主页

如何用BERT做生成任务 解读的文章：《It’s Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners》 [必须要GPT3吗？不，BERT的MLM模型也能小样本学习 - 科学空间 Scientific Spaces](https://kexue.fm/archives/7764) code: https://github.com/bojone/Pattern-Exploiting-Training 这个模式特别适合用来做量表查询症状，现有量表表述+UGC，以此检索ugc中的对应的量表题项 ...

文章来源 SimCSE: Simple Contrastive Learning of Sentence Embeddings [GitHub - princeton-nlp/SimCSE: EMNLP 2021] SimCSE: Simple Contrastive Learning of Sentence Embeddings https://arxiv.org/abs/2104.08821 Introduction Our unsupervised SimCSE simply predicts the input sentence itself with only dropout (Srivastava et al., 2014) used as noise (Figur...

论文：[1907.11692] RoBERTa: A Robustly Optimized BERT Pretraining Approach 代码：GitHub - facebookresearch/fairseq: Facebook AI Research Sequence-to-Sequence Toolkit written in Python. Introduction Motivation: Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have ...

文章来源 文章链接：[1904.09223] ERNIE: Enhanced Representation through Knowledge Integration code: GitHub - PaddlePaddle/ERNIE: The official repository for ERNIE 4.5 and ERNIEKit – its industrial-grade development toolkit based on PaddlePaddle. Introduction Motivation These works do not consider the prior knowledge in the sentence. For example, In t...

l 论文：[1909.11942] ALBERT: A Lite BERT for Self-supervised Learning of Language Representations 代码：https://github.com/google-research/ALBERT Introduction Motivation: 应对GPU/TPU memory limitations and longer training times。this study present two parameter reduction techniques to lower memory consumption and increase the training speed of BERT...

前言 pytorch pip install torch==1.13.0+cu116 torchvision==0.14.0+cu116 torchaudio==0.13.0 -f https://download.pytorch.org/whl/torch_stable.html -i https://pypi.tuna.tsinghua.edu.cn/simple

前言 反刍与反思 反刍与共同反刍 Co-rumination（共同反刍） 指的是一种发生在人际互动中的沟通模式，特点是过度、反复地讨论个人问题以及相关的负面情绪 。它不仅仅是分享烦恼，而是陷入对问题细节和负面感受的持续纠结和猜测。 co-rumination can be described as an interpersonal emotional regulation strategy that is characterized by extensive and repetitive dwelling about problems encouraged by peers (Davidson et al., 2014; Rose, 2002). 它像一把双刃剑，可能增强亲密...

前言 本教程适用于某一领域的初步探索，快速了解该领域当前的研究问题、研究方法、研究结论，这些信息也可以辅助论文撰写时文献综述部分的写作。 该方法不适用于做文献的精读，如果想做文献精读可以用这个工具智谱清言，强烈推荐 检索并下载文献的题录 使用web of science 下载文献的题录信息，（一定要包含摘要），存储格式为.ris 然后导入到zotero。这个步骤相信大多数人都会，就不再赘述 使用AI+zotero插件总结文献 原理： 通过zotero插件调用免费的大模型API，对文献的题录信息（标题、摘要、关键词）进行快速总结，返回表格形式的 研究问题、研究方法、研究结论 step1 下载并安装zotero的GPT插件 这是插件的介绍 Awesome ...