Prompting High-tech Industrialization through Mechanism Innovation

With the further implementation of the knowledge innovation program (KIP), piloted by the Chinese Academy of Sciences (CAS), encouraging progress has been made in contributing to the development of the country's high-tech industry, and forging S&T cooperation with local governments and industrial sectors. This was revealed at the Second CAS Conference on High-tech Industrialization April 25 - 29 in Shenzhen, Guangdong Province.

A Communication Theory Perspective on Prompting Engineering Methods for Large Language Models

The springing up of large language models(LLMs)has shifted the community from single-task-orientated natural language processing(NLP)research to a holistic end-to-end multi-task learning paradigm.Along this line of research endeavors in the area,LLM-based prompting methods have attracted much attention,partially due to the technological advantages brought by prompt engineering(PE)as well as the underlying NLP principles disclosed by various prompting methods.Traditional supervised learning usually requires training a model based on labeled data and then making predictions.In contrast,PE methods directly use the powerful capabilities of existing LLMs(e.g.,GPT-3 and GPT-4)via composing appropriate prompts,especially under few-shot or zero-shot scenarios.Facing the abundance of studies related to the prompting and the ever-evolving nature of this field,this article aims to 1)illustrate a novel perspective to review existing PE methods within the well-established communication theory framework,2)facilitate a better/deeper understanding of developing trends of existing PE methods used in three typical tasks,and 3)shed light on promising research directions for future PE methods.

VAGen:waterbody segmentation with prompting for visual in‑context learning

Effective water management and flood prevention are critical challenges encountered by both urban and rural areas,necessitating precise and prompt monitoring of waterbodies.As a fundamental step in the monitoring process,waterbody segmentation involves precisely delineating waterbody boundaries from imagery.Previous research using satellite images often lacks the resolution and contextual detail needed for local-scale analysis.In response to these challenges,this study seeks to address them by leveraging common natural images that are more easily accessible and provide higher resolution and more contextual information compared to satellite images.However,the segmentation of waterbodies from ordinary images faces several obstacles,including variations in lighting,occlusions from objects like trees and buildings,and reflections on the water surface,all of which can mislead algorithms.Additionally,the diverse shapes and textures of waterbodies,alongside complex backgrounds,further complicate this task.While large-scale vision models have typically been leveraged for their generalizability across various downstream tasks that are pre-trained on large datasets,their application to waterbody segmentation from ground-level images remains underexplored.Hence,this research proposed the Visual Aquatic Generalist(VAGen)as a countermeasure.Being a lightweight model for waterbody segmentation inspired by visual In-Context Learning(ICL)and Visual Prompting(VP),VAGen refines large visual models by innovatively adding learnable perturbations to enhance the quality of prompts in ICL.As demonstrated by the experimental results,VAGen demonstrated a significant increase in the mean Intersection over Union(mIoU)metric,showing a 22.38%enhancement when compared to the baseline model that lacked the integration of learnable prompts.Moreover,VAGen surpassed the current stateof-the-art(SOTA)task-specific models designed for waterbody segmentation by 6.20%.The performance evaluation and analysis of VAGen indicated its capacity to substantially reduce the number of trainable parameters and computational overhead,and proved its feasibility to be deployed on cost-limited devices including unmanned aerial vehicles(UAVs)and mobile computing platforms.This study thereby makes a valuable contribution to the field of computer vision,offering practical solutions for engineering applications related to urban flood monitoring,agricultural water resource management,and environmental conservation efforts.

CLIP-SP:Vision-language model with adaptive prompting for scene parsing

We present a novel framework,CLIPSP,and a novel adaptive prompt method to leverage pre-trained knowledge from CLIP for scene parsing.Our approach addresses the limitations of DenseCLIP,which demonstrates the superior image segmentation provided by CLIP pre-trained models over ImageNet pre-trained models,but struggles with rough pixel-text score maps for complex scene parsing.We argue that,as they contain all textual information in a dataset,the pixel-text score maps,i.e.,dense prompts,are inevitably mixed with noise.To overcome this challenge,we propose a two-step method.Firstly,we extract visual and language features and perform multi-label classification to identify the most likely categories in the input images.Secondly,based on the top-k categories and confidence scores,our method generates scene tokens which can be treated as adaptive prompts for implicit modeling of scenes,and incorporates them into the visual features fed into the decoder for segmentation.Our method imposes a constraint on prompts and suppresses the probability of irrelevant categories appearing in the scene parsing results.Our method achieves competitive performance,limited by the available visual-language pre-trained models.Our CLIP-SP performs 1.14%better(in terms of mIoU)than DenseCLIP on ADE20K,using a ResNet-50 backbone.

基于中国生肖文化基因的IP形象智能生成设计方法研究

目的 为实现中国生肖文化遗产的数字化保护与可持续设计创新,提出基于文化基因分析,以Prompt权重计算调控IP形象呈现的智能生成设计方法。方法 结合深度学习技术与文化基因理论,以层次分析法计算生肖IP形象智能生成设计的Prompt权重,并根据优先级融入Midjourney图像生成过程,通过分组实验进行模糊综合评价检验效果。结果 该方法在智能生成设计中高效、有导向性地调节了生肖IP形象的视觉表征,使其符合文化内涵并具备系列感。结论 在使用智能生成设计工具时,应强调人机协同与专业把控。基于对生肖文化中“主体性”“间体性”与“时代性”基因的分析,创作者能够更精准地对Prompt排序及表述进行优化,以调控生肖IP形象智能生成设计结果,在新时代助力珍贵民俗文化的活态传承、永续发展。

Prompting and Tuning: A Two-Stage Unsupervised Domain Adaptive Person Re-identification Method on Vision Transformer Backbone

This paper explores the Vision Transformer(ViT)backbone for Unsupervised Domain Adaptive(UDA)person Re-Identification(Re-ID).While some recent studies have validated ViT for supervised Re-ID,no study has yet to use ViT for UDA Re-ID.We observe that the ViT structure provides a unique advantage for UDA Re-ID,i.e.,it has a prompt(the learnable class token)at its bottom layer,that can be used to efficiently condition the deep model for the underlying domain.To utilize this advantage,we propose a novel two-stage UDA pipeline named Prompting And Tuning(PAT)which consists of a prompt learning stage and a subsequent fine-tuning stage.In the first stage,PAT roughly adapts the model from source to target domain by learning the prompts for two domains,while in the second stage,PAT fine-tunes the entire backbone for further adaption to increase the accuracy.Although these two stages both adopt the pseudo labels for training,we show that they have different data preferences.With these two preferences,prompt learning and fine-tuning integrated well with each other and jointly facilitated a competitive PAT method for UDA Re-ID.

Compositional Prompting Video-language Models to Understand Procedure in Instructional Videos

Instructional videos are very useful for completing complex daily tasks,which naturally contain abundant clip-narration pairs.Existing works for procedure understanding are keen on pretraining various video-language models with these pairs and then finetuning downstream classifiers and localizers in predetermined category space.These video-language models are proficient at representing short-term actions,basic objects,and their combinations,but they are still far from understanding long-term procedures.In addition,the predetermined procedure category faces the problem of combination disaster and is inherently inapt to unseen procedures.Therefore,we propose a novel compositional prompt learning(CPL)framework to understand long-term procedures by prompting short-term video-language models and reformulating several classical procedure understanding tasks into general video-text matching problems.Specifically,the proposed CPL consists of one visual prompt and three compositional textual prompts(including the action prompt,object prompt,and procedure prompt),which could compositionally distill knowledge from short-term video-language models to facilitate long-term procedure understanding.Besides,the task reformulation enables our CPL to perform well in all zero-shot,few-shot,and fully-supervised settings.Extensive experiments on two widely-used datasets for procedure understanding demonstrate the effectiveness of the proposed approach.

基于完形填空的方面级情感四元组预测

方面情感四元组预测(ASQP)任务旨在从给定的评论语句中提取所有方面词以及相应的方面类别、观点表达和情感极性,有助于全面了解用户对产品或服务不同方面的评价情况.现有情感四元组预测方法主要存在以下局限:(1)判别式模型没有利用prompt捕获情感元素之间的语义关系;(2)生成式模型要么简单地将情感元素类型标签组合形成prompt,缺乏理解标签类型涵义的语境;要么将离散模板作为解码器的输入,而编码器则无法捕获到模板中情感元素之间的语义关系.为了缓解这些问题,本文首先基于完形填空思想研制离散和连续2类prompt,提供理解4个情感元素类型涵义的语境,帮助模型更好地捕获情感元素之间的语义关系;然后,基于设计的prompt,提出C-ASQP框架,包含判别式模型DC-ASQP和生成式模型GC-ASQP.在DC-ASQP中,采用2阶段策略,先预测4个情感元素中2个较为容易的情感元素,再将预测结果嵌入到设计的prompt中,帮助模型理解情感元素类型的涵义,从而有效预测另外2个情感元素.在GC-ASQP中,将设计的prompt作为编码器的输入,借助预训练模型的学习模式,充分利用预训练模型蕴含的知识提升四元组的生成效果.实验结果显示,DC-ASQP模型在4个常用数据集上的F1值相比同类判别式最优模型分别提高4.70%、6.48%、6.97%和2.60%,GC-ASQP模型的F1值比最优基准模型分别提高0.86%、1.67%、0.15%和1.02%,验证了将ASQP建模为完形填空任务的有效性,所设计的2类prompt以及C-ASQP框架是有效的.

基于RoBERTa和T5的两阶段医学术语标准化

医学术语标准化作为消除实体歧义性的重要手段,被广泛应用于知识图谱的构建过程之中.针对医学领域涉及大量的专业术语和复杂的表述方式,传统匹配模型往往难以达到较高的准确率的问题,提出语义召回加精准排序的两阶段模型来提升医学术语标准化效果.首先在语义召回阶段基于改进的有监督对比学习和RoBERTa-wwm提出语义表征模型CL-BERT,通过CL-BERT生成实体的语义表征向量,根据向量之间的余弦相似度进行召回并得到标准词候选集,其次在精准排序阶段使用T5结合prompt tuning构建语义精准匹配模型,并将FGM对抗训练应用到模型训练中,然后使用精准匹配模型对原词和标准词候选集分别进行精准排序得到最终标准词.采用ccks2019公开数据集进行实验,F1值达到了0.9206,实验结果表明所提出的两阶段模型具有较高的性能,为实现医学术语标准化提供了新思路.

自然语言处理技术视角下的编程演进研究

人类想要与计算机沟通就对编程语言的语法和结构有一定的了解。随着自然语言处理技术的发展,让计算机听懂人类语言不再是梦想。ChatGPT的应用催生了Prompt(提示工程),Prompt可以直接使用自然语言表达用户的意图,大大降低了编程的门槛。文章从编程语言发展历程,了解Prompt原理和应用,在自然语言处理时代,编程的边界在扩大。

基于Prompt的两阶段澄清问题生成方法

在自然语言相关系统中,当用户输入存在歧义时,生成澄清问题询问用户有助于系统理解用户需求;基于Prompt的方法可以更好地挖掘预训练语言模型的潜在知识,但往往需要手动设计模板,限制其生成澄清问题的多样性。为解决这一问题,提出了TSCQG(two-stage clarification question generation)方法。首先,在动态Prompt模板生成阶段,利用歧义上下文和预训练语言模型生成动态的Prompt模板;然后在缺失信息生成阶段,将Prompt模板与外部知识相结合,充分利用预训练语言模型的生成能力生成相应的缺失信息。实验结果表明,在CLAQUA数据集的多轮对话情况中,BLEU值和ROUGE-L值分别达到了58.31和84.33,在ClariQ-FKw数据集上,BLEU值和ROUGE-L值分别达到了31.18和58.86。实验结果证明了TSCQG方法在澄清问题生成任务上的有效性。

知识图谱和大语言模型在航天测控问答系统中的融合应用研究

随着自然语言处理技术的发展,智能化检索与问答系统不断发展,为改善传统知识图谱语义解析能力不足和当前通用大语言模型在垂直领域对知识学习不够深入的问题,提出了融合知识图谱的大语言模型方法,进行了两步优化:首先,利用知识图谱在命名实体识别和关系抽取的基础上,构建大模型提示Prompt模板,进行辅助增强生成,利用图谱中存储的数据提供相关来源;其次,利用低阶适应性微调(LoRA)策略冻结大模型原有参数,增加部分网络参数进行微调训练,优化模型在航天测控领域的知识储备与理解。通过两步改进提高了模型整体在语义解析和知识细节上的理解掌握,结合航天测控领域的相关教材、报告和手册等资料,搭建了知识问答系统,取得了较好的效果,说明了该方法具有一定的应用价值。

Social Radars: Finding Targets in Cyberspace for Cybersecurity

INSPIRED by the insight from American political scientist Lasswell, who summarized the environmental role in societal surveillance [1], Schramm coined the term “social radar” [2] as it resembles the activities of radar in collecting and processing information, playing a crucial role in helping humans perceive changes in the internal and external environment and promptly adjusting adaptive behaviors.

Considerations for exerkine research focusing on the response to exercise training

García-Hermoso and colleagues1 recently published a systematic literature review and meta-analysis on exercise training-induced changes in exerkine concentrations in type 2diabetes mellitus patients,providing a contemporary view on how exerkines respond to exercise training.That review prompted us to highlight 2 additional considerations that should be taken into account when studying the response of exerkines to exercise training.Firstly,whether exerkines can exhibit discordant responses to acute exercise compared to exercise training,and secondly,the need to consider the residual effects of the most recent exercise bout.

新闻生产中生成式AI技术的应用研究

随着人工智能技术的快速发展,生成式AI在文本生成、视觉内容制作、音乐创作、视频编辑等方面展现出巨大的潜力,能够从多个维度优化新闻内容的表现形式,如文本、视觉、音乐、视频等。文章主要探讨生成式AI技术在新闻生产中的应用,旨在分析如何应用该技术提升新闻制作效率和质量,同时探讨其在文本、视觉、音乐和视频内容创作中的具体应用场景。文章采用文献综述与案例分析相结合的方法,通过分析生成式AI技术及其具体应用案例,探讨其在新闻生产中的实际应用效果,并结合实践操作,深入探讨生成式AI技术在新闻生产中的四种主要应用策略,包括Prompt应用策略、RAG应用策略、微调应用策略和重新训练应用策略。研究结果表明,在新闻文本生产中,生成式AI可以自动生成新闻稿件,从而减轻记者的工作负担。在新闻视觉内容制作中,生成式AI可以生成高质量的图像和视频,并快速生成短视频内容。在新闻音乐创作中,生成式AI可以根据新闻主题生成背景音乐,增强新闻的情感感染力。然而,文章也指出,生成式AI技术的应用存在局限性,需要在实际操作中进一步优化技术模型,以确保生成内容的准确可靠。文章旨在为新闻行业提供一种新的内容生产方式,推动新闻生产的技术革新。

The immune system:uncharted pathways between senses and the brain

Humans and animals use the classic five senses-sight,hearing,touch,smell,and taste-to detect and monitor their environment,with the sense of position and movement often referred to as the sixth sense.The perception of external signals through the senses is essential to an organism's survival,transmitting signals to the central nervous system(CNS)and prompting physiological changes in other biological systems.In addition to the direct effects of sense-induced mediators in the brain.

Comprehension-driven design of advanced multi-block single-ion conducting polymer electrolytes for high-performance lithium-metal batteries

The continuously growing importance of batteries for powering(hybrid)electric vehicles and storing renewable energy has prompted a renewed focus on lithium-metal batteries(LMBs)in recent years,as its high theoretical specific capacity of about 3860 mA h g^(-1) and very low redox potential(-3.04 V vs.the standard hydrogen electrode)promise substantially higher energy densities compared to current lithium-ion batteries(LIBs)[1].However,lithium metal electrodes face severe challenges associated with the risk of dendritic lithium deposition and the high reactivity with traditional organic liquid electrolytes,resulting in a continuous loss of electrochemically active lithium and a relatively low Coulombic efficiency[2].To address these challenges,solid inorganic and polymer electrolytes have emerged as a potentially saferalternative.

Sinensetin protects against periodontitis through binding to Bach1 enhancing its ubiquitination degradation and improving oxidative stress

Periodontitis is a chronic inflammatory and immune reactive disease induced by the subgingival biofilm.The therapeutic effect for susceptible patients is often unsatisfactory due to excessive inflammatory response and oxidative stress.Sinensetin(Sin)is a nature polymethoxylated flavonoid with anti-inflammatory and antioxidant activities.Our study aimed to explore the beneficial effect of Sin on periodontitis and the specific molecular mechanisms.We found that Sin attenuated oxidative stress and inflammatory levels of periodontal ligament cells(PDLCs)under inflammatory conditions.Administered Sin to rats with ligation-induced periodontitis models exhibited a protective effect against periodontitis in vivo.By molecular docking,we identified Bach1 as a strong binding target of Sin,and this binding was further verified by cellular thermal displacement assay and immunofluorescence assays.Chromatin immunoprecipitation-quantitative polymerase chain reaction results also revealed that Sin obstructed the binding of Bach1 to the HMOX1 promoter,subsequently upregulating the expression of the key antioxidant factor HO-1.Further functional experiments with Bach1 knocked down and overexpressed verified Bach1 as a key target for Sin to exert its antioxidant effects.Additionally,we demonstrated that Sin prompted the reduction of Bach1 by potentiating the ubiquitination degradation of Bach1,thereby inducing HO-1 expression and inhibiting oxidative stress.Overall,Sin could be a promising drug candidate for the treatment of periodontitis by targeting binding to Bach1.

Human radiological safety assessment for petawatt laser-driven ion acceleration experiments in CLAPA-T

The newly built Compact Laser Plasma Accelerator-Therapy facility at Peking University will deliver 60 J/1 Hz laser pulses with 30 fs duration.Driven by this petawatt laser facility,proton beams with energy up to 200 MeV are expected to be generated for tumor therapy.During high-repetition operation,both prompt radiation and residual radiation may cause safety problems.Therefore,human radiological safety assessment before commissioning is essential.In this paper,we simulate both prompt and residual radiation using the Geant4 and FLUKA Monte Carlo codes with reasonable proton and as-produced electron beam parameters.We find that the prompt radiation can be shielded well by the concrete wall of the experimental hall,but the risk from residual radiation is nonnegligible and necessitates adequate radiation cooling.On the basis of the simulation results,we discuss the constraints imposed by radiation safety considerations on the annual working time,and we propose radiation cooling strategies for different shooting modes.

Transactions of Nanjing University of Aeronautics and Astronautics Information for Contributors

Transactions of Nanjing University of Aeronautics&Astronautics (TNUAA) is a bimonthly journal facing international academic community.Emphasizing prompt and effective dissemination of key data and new scientific insights,TNUAA publishes experimental and theoretical papers bearing on applications to all branches of aeronautics, astronautics and civil aviation.