This article presents a comprehensive overview of search term recommendation in QQ Browser, covering various recommendation scenarios, challenges, query library architecture, multi‑task ranking models, coarse‑to‑fine ranking pipelines, auto‑completion strategies, and future research directions.
An in‑depth look at MonaLisa, an AI‑driven visual UI inspection platform that automates pixel‑level checks by converting design files to HTML, matching DOM structures, and leveraging OpenCV and MobileNetV2 to generate reports, reduce manual rework, and streamline front‑end development workflows.
This presentation explores how deep semantic role labeling and multimodal data fusion of biomedical literature can build fine‑grained knowledge graphs that enable drug repurposing, describing the AGAC corpus, modeling approaches, evaluation results, and future research directions.
This article introduces event graphs as a knowledge‑centric representation of dynamic events, explains their construction and real‑time processing in Meituan's intelligent customer service, and demonstrates applications such as event timeline extraction, hotspot detection, event prediction, multi‑turn dialogue guidance, and business decision support.
The article compares Google’s Imagen and OpenAI’s DALL·E by feeding them the whimsical "Tiger Wearing VR" prompt, showcasing each model’s visual style, underlying architecture—including CLIP, diffusion, and T5‑XXL—and community reactions to the resulting AI‑generated artwork.
This article details Youku's end‑side solution for real‑time danmaku‑through‑human rendering, covering its background, value, client‑side advantages, technical challenges, OPR architecture, processing pipeline, engineering optimizations, test results, and future outlook for mobile intelligent rendering.
From its quirky start as a teenage iPhone chatbot to becoming the central hub for open‑source transformer models, Hugging Face has grown into a fast‑rising AI platform, securing $100 million Series C funding, serving thousands of organizations, and aiming to democratize machine learning.
This article provides a detailed overview of large‑scale RDMA technology, covering basic concepts, major protocols, network‑level techniques such as congestion control, lossless‑to‑lossy evolution and multipath, virtualization, communication libraries for AI training and storage, performance tuning, monitoring, and real‑world deployment experiences.
This article explains Meituan's 'Meituan Brain' initiative, detailing the construction of life‑service knowledge graphs—including tag and dish graphs—through data mining, semantic extraction, synonym discovery, graph labeling, and applications such as open QA, search ranking, and recommendation using AI and GNN techniques.
Knowledge Distillation, a teacher‑student model compression technique, enables large, high‑performing deep neural networks to transfer their learned representations to smaller models, achieving comparable accuracy with faster inference, reduced resource consumption, and broader applicability in computer‑vision tasks.
Alibaba Mama’s advertising engine has been transformed into a serverless, cloud‑native platform that unifies runtime, data, and business abstractions, adopts vector‑ and model‑based recall with offline pre‑computed pipelines, implements multi‑stage AI‑driven bidding and auction mechanisms, and leverages large‑scale generative AI for creative assets, thereby accelerating feature rollout, cutting latency, and boosting merchant value.
This tutorial demonstrates how to use Huawei Cloud ModelArts and the Proximal Policy Optimization (PPO) reinforcement‑learning algorithm to train an AI agent that can clear most Super Mario levels within about 1500 episodes, even for users with no coding experience.
This article presents Kuaishou's large‑scale recommendation system, detailing its precision‑ranking architecture, the PPNet CTR model, multi‑domain multi‑task learning, short‑ and long‑term behavior sequence modeling, billion‑feature trillion‑parameter engineering, and future research directions.
Youku’s Digital Media Lab built an AI‑powered restoration pipeline that uses multi‑frame super‑resolution, denoising, deblocking, deblurring and GAN‑based detail generation to automatically revive classic animated films, removing noise and artifacts while preserving fine lines, enabling high‑definition viewing for modern audiences.
This talk presents JD's intelligent creative ecosystem for advertising, detailing the construction of a content understanding system, AI-driven content generation (including OCR, image tagging, video summarization, and copy generation), and multimodal creative selection and distribution, highlighting challenges, solutions, and business impact.
Harbor’s growth engine combines a passive, attribution‑driven traffic‑allocation algorithm with componentized ranking, search, and marketing systems—using pairwise/Listwise models, multi‑task CTR/CVR prediction, and automated strategy triggers—to align short‑term efficiency with long‑term LTV goals while moving toward causal inference and domain‑expert‑driven general models.
This tutorial explains how to use the open‑source PaddleGAN Wav2Lip model to synchronise any face or avatar with arbitrary speech, covering the underlying AI principles, required installations, and step‑by‑step command‑line usage for creating high‑quality dubbing videos.
This article showcases five cutting‑edge open‑source projects—from a ROS‑based system that lets a gamepad animate facial muscles, to AI‑driven video inpainting, text‑to‑image generation, eye‑gaze computer control, and a comprehensive Wi‑Fi cracking toolkit—each pushing the boundaries of modern tech.
This article explains how a multimodal deep learning model combines static and temporal game data to predict frame rates, helping identify performance bottlenecks and improve client smoothness through feature fusion, data pipelines, and real‑time inference in modern games.
This article details Meituan's journey from rapid database growth and operational bottlenecks to a multi‑year roadmap that combines platform‑level monitoring, rule‑based and AI‑enhanced root‑cause analysis, and automated remediation, ultimately delivering measurable improvements in alert accuracy, recall rates, and overall database reliability.
This article presents a comprehensive overview of how contrastive learning can alleviate data sparsity and distribution bias in recommendation systems, detailing its theoretical advantages, recent research progress in computer vision and NLP, and a multi‑task self‑supervised framework applied to Kuaishou's short‑video ranking pipeline with significant offline and online performance gains.
Alibaba Mama boosts external-media advertising ROI by deploying AI-driven models—privacy-preserving federated learning, hierarchical representation integration, uncertainty-regularized knowledge distillation, and calibrated DNNs—to overcome missing user-preference data, sparse post-click conversions, sample-selection bias, and probability-calibration challenges.
This article presents a comprehensive overview of the challenges in e‑commerce product copy generation, introduces a unified framework comprising a copy generation system, a copy‑cleaning subsystem, and a quality evaluation module, and details practical optimization techniques applied to short and long copy scenarios.
The article examines how GPUs have evolved from gaming accelerators to essential data‑center processors for AI, HPC, and scientific workloads, and compares the latest server‑grade offerings from Nvidia, AMD, and Intel—including performance specs, memory technologies, interconnects, and software ecosystems—highlighting the fierce competition shaping the future of compute.
This article reviews Huawei Noah's Ark Lab's work on modern information‑flow recommendation, covering the evolution from collaborative filtering to deep learning, the application of BERT‑based pre‑training for news ranking, multimodal user‑interface modeling, practical deployment challenges, and future research directions.
This article presents a complete Python implementation of the classic Gomoku board game with Pygame, explaining the rules, showing a dynamic demo, providing the full source code for the checkerboard logic, AI opponent, rendering, and instructions for running the game on your own machine.
This article presents a multimodal approach that combines dynamic analysis and graph machine learning to generate and apply social relationship graphs in videos, detailing problem background, graph generation modules, applications such as video retrieval, experimental results, and future research directions.
This article debunks the overused term “intuitive design,” explains what intuition really means, shows how familiarity—not unconscious reasoning—drives perceived intuitiveness, and demonstrates how applying proven design patterns and emerging AI technologies can create clearer, more efficient user experiences.
In a TVP forum hosted by Li Kaifu and Shen Chunhua, experts trace AI’s 70‑year journey, discuss the origins of the book “AI Future in Progress,” analyze investment stages, AI‑cloud synergy, NLP breakthroughs, medical applications, societal impacts, data privacy, and the challenges facing traditional enterprises.
This article describes the motivation, micro‑service architecture, code implementation, V1.0 browser‑based NLP prototype, V2.0 AI‑enhanced version with BM25 and BERT, integration with ChatUI, DingTalk bot, command‑based automation, and future plans for the internal intelligent QA chatbot named Jarvis.
The article provides a comprehensive English recap of Google I/O 2022, covering AI enhancements across Google products, new Android 13 privacy features, expanded Jetpack libraries and Compose improvements, as well as Flutter 3's multi‑platform support and related ecosystem developments.
This article presents Meituan's large‑scale knowledge graph, its integration into location‑based recommendation, the challenges of explainability, domain diversity, data sparsity and spatiotemporal complexity, and describes a dual‑memory neural network and cross‑domain learning approach that improve recall, ranking and recommendation fairness.
The 2022 Beijing University Smart Interaction Design Forum gathered industry leaders and academics to explore AI‑driven interaction design, covering digital assistants, visual search, automotive HMI, and multimodal virtual humans, and emphasized the growing synergy between research and practice.
The 4th International Workshop on Deep Learning Practice for High‑Dimensional Sparse and Imbalanced Data (DLP‑KDD 2022) invites submissions on deep‑learning systems, data representation, and user modeling for large‑scale sparse data, with a deadline of May 26, 2022 and acceptance notifications by June 20, 2022.
SAS announced native Python support in its analytics platform, offering code examples, cloud‑native AI capabilities, and seamless integration with SAS Studio and other SAS products to empower data scientists to develop and deploy Python models alongside SAS code.
The article examines the challenges advertisers face in a monopolized traffic environment, presents a three‑layer audience insight system, explores graph‑based relationship network methods, and discusses various modeling approaches—including single‑stage, cross‑stage, and multi‑task learning—along with practical considerations for data security and platform integration.
This article reviews recent advances in knowledge graph reasoning, introduces the RNNLogic framework that jointly learns a rule‑generating LSTM and a stochastic logic programming predictor, and demonstrates its competitive performance and interpretability on benchmark datasets while outlining future neural‑symbolic directions.
This article presents a comprehensive overview of an intelligent risk control middle‑platform, covering its design background, the five‑characteristics and "five‑all double‑core" concept, full‑cycle strategy and model lifecycle management, business architecture, and real‑world application cases, highlighting the integration of rule‑based and AI‑driven decision engines.
This article explains the background, evaluation metrics, and technical framework of computer‑assisted language learning, detailing how modern AI‑driven speech assessment systems use HMM‑DNN models, GOP scoring, conformer architectures, streaming solutions, and edge‑cloud deployment to deliver accurate, low‑latency pronunciation feedback for learners.
This tutorial explains what Faiss is, how to install it, construct various indexes such as IndexFlatL2, IndexIVFFlat, and IndexIVFPQ, and demonstrates code examples for building and querying vector similarity search pipelines while discussing speed‑accuracy trade‑offs.
This talk presents Meituan Brain's massive product knowledge graph, explains how knowledge‑enhanced models improve product title parsing, category association and sample governance, and demonstrates the resulting gains in search, recommendation and other downstream services while keeping the system online‑controllable and explainable.
Tencent’s Hunyuan AI large model shattered records by scoring 80.888 to claim first place on the CLUE benchmark, showcasing its advanced natural language processing, multimodal abilities, curriculum‑learning training approach, and real‑world deployments in WeChat Search and advertising.
The AI Engineering Online Forum, co‑hosted by China Academy of Information and Communications Technology, unveiled the industry’s first AI Model Development and Management (Model/MLOps) maturity standard, featured expert insights from finance, telecom, and tech leaders, and showcased practical MLOps implementations across banking, Huawei, and AI startups.
This article presents open‑source AI solutions for financial form recognition and insurance smart Q&A, detailing the challenges, model choices, optimization strategies, performance results, and deployment methods using PaddleOCR, PaddleNLP, LayoutXLM, RocketQA and SimCSE.
The article explores the rapid growth of privacy computing in China, highlighting new data protection laws, a surge in patents, real‑world applications such as medical data sharing, and personal stories of Ant Group engineers who develop cutting‑edge cryptographic frameworks and champion interdisciplinary research.
The article explains the drawbacks of the traditional generalized second‑price auction for online ads, proposes monitoring and probabilistically throttling fast‑spending campaigns using a time‑window based budget pacing model with a Pass‑Through Rate (PTR), and references the related research paper while also offering free resource links.
Tencent’s newly unveiled HunYuan AI model achieved a grand‑slam by ranking first on the five most authoritative cross‑modal video retrieval datasets, showcasing a hierarchical multimodal approach that dramatically boosts retrieval precision and promises broad impact for both research and industry applications.
Microsoft's Jigsaw tool leverages post‑processing and user feedback to improve large language model code synthesis, especially for Python Pandas, achieving up to 80% accuracy and aiming to automate code verification and debugging for developers.
Volcano, the first cloud‑native batch computing project contributed by Huawei Cloud, has been promoted to a CNCF incubating project, signaling broad industry recognition and outlining a roadmap that includes cross‑cloud scheduling, GPU virtualization, fine‑grained resource management, and workflow orchestration for AI and big‑data workloads.
This article presents a comprehensive overview of an intelligent risk control platform, covering its design background, six core characteristics, the "five‑full double‑core" concept, end‑to‑end strategy and model lifecycle management, business architecture atomization, and real‑world anti‑fraud case studies.
This article reviews DeepSpeed‑MoE, an end‑to‑end system that introduces new MoE architectures, model‑compression techniques, and highly optimized inference pipelines, detailing its motivation, design of PR‑MoE (Pyramid‑MoE and Residual‑MoE), distributed parallel strategies, communication and kernel optimizations, and performance gains over dense baselines.
The article outlines DeWu’s warehouse‑management recommendation system, which combines an online‑near‑line‑offline architecture to quickly recall viable shelf slots and rank them by space utilization, travel time, and sales potential, enabling automated, constraint‑aware placement that cuts picking time and inventory costs.
The article examines the evolution of project management standards, highlights China’s recent national guidelines, and presents JD.com’s intelligent shopping guide AI project as a case study illustrating how business‑driven value streams, advanced technologies and rigorous target‑benefit management can shape modern project management practices.
Effective e‑commerce supply‑demand coordination requires accurate SKU‑level forecasting, optimized replenishment under MOQ and lead‑time constraints, and dynamic post‑sale traffic control, using a blend of time‑series, tree‑based and deep learning models together with expert knowledge to minimize inventory costs and avoid stock‑outs.
The 2022 CCKS conference invites submissions of open‑resource papers on knowledge graphs and semantic computing, offering details on submission deadlines, required paper length, evaluation criteria, example resources, chairs, important dates, and links to templates and the EasyChair submission portal.
This article presents Hello Mobility’s precise marketing system, covering its business background, value, framework, algorithmic capabilities such as Pu‑Learning LookAlike modeling, TSA semi‑supervised learning, and Graph Embedding, as well as identified pain points, project impact, and future directions for scaling and automation.
This tutorial shows how to use the AnimeGAN v2 model to convert personal photos into comic‑style avatars, expose the AI service via a Python Bottle API on a Serverless platform, configure deployment with Serverless Devs, and integrate the backend into a WeChat mini‑program for end‑users.
This article presents a detailed case study of using AI speech‑recognition techniques—including acoustic modeling with VGG, pypinyin conversion, feature extraction, and CTC decoding—to automatically verify game dialogue audio against script text, outlining the workflow, challenges, implementation details, and experimental results.
This article examines JD Logistics' integrated digital supply chain, detailing its evolution, the construction of a multi‑stage intelligent planning system, algorithm and engineering platforms, digital twin applications, real‑world case studies, and insights into future talent and ecosystem development.
Jack Cui presents three recently open‑sourced CVPR 2022 algorithms—DualStyleGAN for high‑resolution portrait style transfer, RQ‑VAE Transformer for improved text‑to‑image generation, and VFD for deep‑fake detection—detailing their functionality, usage options, and providing links to code repositories and demo platforms.
This article presents the evolution of Alibaba's display advertising auction mechanisms, introducing the AIDA decision‑allocation framework, the Deep GSP multi‑objective smart auction, and the end‑to‑end Neural Auction, while discussing their economic theory, engineering platformization, business impact, and future research directions.
This article describes the end‑to‑end design, architecture, code implementation, and deployment steps for an internal intelligent QA chatbot named “Jarvis”, covering its V1.0 browser‑based prototype, V2.0 AI‑enhanced version, DingTalk integration, automation features, and future roadmap.
The recall module in recommendation systems retrieves thousands of items from massive pools using parallel non-personalized and personalized paths—such as hot-item, content-based, behavior-based, and deep-model recall—prioritizing coverage and low latency while addressing challenges like hard-negative sampling, selection bias, objective alignment, and channel competition to feed downstream ranking.
This article explains what artificial intelligence is, distinguishes weak and strong AI, explores its applications in games, product testing, and NetEase's Fuxi platform, and introduces fundamental machine‑learning concepts such as supervised, unsupervised, and reinforcement learning, as well as neural networks and loss functions.
This presentation details how Hallo Mobility tackles the challenges of information‑flow ad modeling by describing the ad ecosystem, the company’s business evolution, and the advertiser‑side algorithmic solutions—including plan‑level quality detection, creative‑level uplift modeling, feature‑cross engineering, and pre‑bid user screening—while outlining future directions for automated, data‑driven ad delivery.
On March 29, the Beijing FinTech Industry Alliance published two white‑papers—‘Multi‑Party Secure Computation Financial Application Status and Implementation Guide’ and ‘Federated Learning Technology Financial Application White Paper’—detailing policies, standards, case studies, and recommendations for deploying privacy‑preserving AI technologies in the financial sector.
This article introduces the AIDA framework for Alibaba's display advertising, detailing the business background, multi‑objective optimization challenges, the design of Deep GSP and Neural Auction mechanisms powered by deep learning and reinforcement learning, and outlines future technical and platform directions while also announcing recruitment opportunities.
This article presents a technical overview of the Youdao Dictionary Pen, describing its hardware design, real‑time scanning and point‑query image processing, on‑device offline translation with model compression techniques, and the high‑performance Edge ML Library (EMLL) that enables efficient AI inference on constrained edge hardware.
This article presents Kuaishou's graph‑based item‑to‑item (I2I) recall pipeline for short‑video recommendation, detailing the business challenges, pipeline architecture, optimization techniques such as similarity‑measure tricks, graph structure learning, edge‑weight learning, and future research directions.
This article presents a comprehensive overview of Huya's live‑streaming recommendation system, covering business background, overall architecture, vector‑based retrieval, detailed ranking pipeline, technical challenges, deployment strategies, scalability, and future outlook.
Cloud computing delivers on‑demand, pay‑as‑you‑go resources through IaaS, PaaS, and SaaS layers, while edge computing extends these services toward devices to meet latency, bandwidth, and privacy demands, using lightweight K3s‑based clusters that address management, connectivity, and resource constraints for applications such as autonomous driving.
The article details Meituan's five‑year journey building a high‑concurrency, AI‑enhanced instant logistics platform, describing its distributed architecture, scalability and reliability practices, fault‑tolerance mechanisms, and future challenges in microservice and unit‑based operations.
The 2021 Taobao Push project introduced an AI‑driven copy‑generation platform that combines template extraction and fine‑tuned Unilm models with diverse beam search, creating diverse, high‑quality push messages, cutting manual costs, and delivering a 10 % click‑through lift and higher material adoption.
This article presents a comprehensive overview of travel intent prediction at Alibaba's Fliggy platform, detailing the unique challenges of low‑frequency travel behavior, multi‑source user actions, a multi‑CNN and time‑attention model design, and experimental evaluations that demonstrate improved recommendation performance.
The article examines how the rapid growth of industrial internet data demands robust privacy‑preserving technologies, outlines the regulatory backdrop, describes the technical challenges of achieving high performance, stability, low cost, flexibility and strong security, and proposes trusted privacy computing as a comprehensive solution for the emerging data‑confidentiality era.
This article explains how to build a Python-based snake game with both human and AI players, detailing the development environment, core features, AI path‑finding logic using BFS, and providing complete source code for a fully functional AI snake that can compete against a human‑controlled snake.
This article explains how the Milvus vector database and the Towhee embedding framework together enable large‑scale, high‑throughput semantic analysis and retrieval of unstructured data such as images, video, and audio by leveraging AI‑powered vectorization and search pipelines.
This article analyzes ten fundamental questions about digital twins, covering definitions, stakeholders, global interest, relationship with smart manufacturing, integration with New IT, scientific challenges, standards, and commercial tools, aiming to guide researchers, policymakers, and practitioners in understanding and applying digital twin technology.
The article surveys digital‑human technology—covering visual form, motion and AI‑driven intelligence—its fast‑growing market, a three‑layer solution stack (hardware/software base, AI platform, application services), end‑to‑end avatar creation workflow, rendering and animation techniques, web‑deployment challenges, and future prospects such as deeper AI, XR/6G and metaverse integration.
This article introduces five emerging programming languages—Bhai-lang, Peregrine, Flix, Skip, and PolyCoder—detailing their origins, unique syntax rules, core features, and where to find their source code or playgrounds for developers seeking fresh, experimental tools.
At the recent GTC event, NVIDIA introduced the H100 GPU built on the Hopper architecture using TSMC 4nm process, featuring 800 billion transistors, 16,896 CUDA cores, up to 700 W power, 3 TB/s memory bandwidth, and a specialized Transformer engine that accelerates large‑model training up to six times faster, alongside the Grace CPU Superchip and new AI supercomputing systems.
The paper reports a year‑long evolution of Alibaba’s search‑advertising keyword recall, replacing the traditional two‑stage rewrite‑and‑score pipeline with a low‑storage binary retrieval model and then a joint recall framework built on a hierarchical bidding graph, delivering near‑full‑precision recall, 16× memory savings, and quota‑free global ranking.
This article explains how Kuaishou leverages graph neural networks for item‑to‑item (I2I) recall in short‑video recommendation, detailing the system background, pipeline architecture, optimization techniques such as similarity measurement, graph structure learning, edge‑weight learning, and future research directions.
The article explains how rapid internet growth created information overload, leading to concise summary services, and how recent AI advances—especially large language models like GPT‑3—enable platforms such as JD.com to automatically generate high‑quality, multimodal product copy that drives sales and supports diverse creative tasks.
This article presents a scenario‑driven methodology for scaling intelligent testing through the five core testing phases—input, execution, analysis, localization, and evaluation—detailing how AI techniques enhance coverage, efficiency, fault detection, and risk assessment.
Tencent’s 2021 R&D Report details a 41% rise in engineering staff, 32 billion new code lines, Go becoming the top language, massive growth in open‑source contributions, breakthroughs in cloud OS, databases, AI, and a commitment to carbon‑neutral technology‑driven social impact.
This talk presents a comprehensive study of invalid query detection in voice assistants, covering the definition and taxonomy of invalid queries, challenges of non‑human interaction and ambiguous intent recognition, data collection and labeling strategies, feature engineering, deep neural network modeling, experimental results, user‑feedback loops, and current performance limits.
This article presents Beike's experience in intelligent real‑estate advertising, detailing the business background, key technical challenges such as conversion‑rate estimation, delayed‑feedback modeling, GEO targeting, and budget allocation, and sharing practical solutions that improved conversion rates by over 10%.
This article details how QQ Music tackled rapid growth in recommendation traffic by redesigning its recommendation architecture, introducing a cloud‑native machine‑learning platform, optimizing data services, and adopting a DAG‑based recall system to improve scalability, flexibility, and development efficiency.
This article explains how semi‑supervised pseudo‑label learning can dramatically improve model accuracy by using a tiny five‑line code snippet that generates pseudo‑labels for unlabeled data, retrains a second model, and avoids data leakage with a proper validation set.
Meituan’s instant logistics platform evolved from vertical services to a micro‑service, distributed architecture that handles massive order‑rider matching, ultra‑low latency, and high availability, leveraging AI for pricing, ETA, scheduling, and employing robust scaling, consistency, and disaster‑recovery techniques.
Carnegie Mellon researchers introduced PolyCoder, a 27‑billion‑parameter open‑source code generation model built on GPT‑2, trained on 249 GB of multi‑language code and achieving superior performance to Codex in C while remaining competitive across eleven other programming languages.
The 2021 software industry review highlights the rise of hybrid work, the continued dominance of microservices, emerging data engineering and AI/ML practices, ethical and sustainability concerns, multi‑cloud and cloud‑native adoption, and anticipates further developments in these areas throughout 2022.
The article surveys deep‑learning‑based face‑recognition systems, detailing detection, preprocessing, and recognition pipelines, describing evaluation metrics such as TAR, FAR, and Rank‑K, reviewing major datasets like LFW, MS‑Celeb‑1M and VGGFace2, and comparing leading architectures—including FaceNet, CenterLoss, SphereFace and InsightFace—while highlighting their strengths, limitations, real‑world applications, and seminal research references.
The JD Intelligent Customer Service team’s NR‑Rino model topped the DROP leaderboard at 90.26% accuracy by enhancing multi‑head predictor architecture and training strategies, showcasing advanced discrete reasoning for machine reading comprehension and promising broader AI applications in finance, logistics, and health.
The Kuaishou audio‑video team designed a high‑precision, low‑latency intelligent danmu‑blocking system that uses advanced image‑segmentation and temporal‑stability techniques to generate accurate masks, improve scene robustness, eliminate mask delay, and enhance user experience across diverse video content.
This talk explores how causal inference techniques are used to identify and mitigate various biases in Kuaishou's recommendation pipeline, covering background theory, recent research advances, practical implementations for popularity and video completion debiasing, and reflections on challenges and future directions.
To prevent dense on‑screen comments from obscuring key video content, Kuaishou’s audio‑video team built a high‑precision, low‑latency intelligent danmu‑blocking system that uses advanced image‑segmentation masks, temporal stability enhancements, SSIM‑based scene detection, and a large‑scale annotated dataset to ensure robust, real‑time protection across diverse video scenarios.
This article explains CLIP’s multimodal contrastive pre‑training, its simple yet effective architecture, code implementation, and how its zero‑shot capability can surpass supervised ImageNet models by leveraging a 400‑million image‑text dataset and shared semantic embeddings.
The article provides a comprehensive analysis of China's enterprise SSD market in 2021, detailing six business models, key component manufacturers, NAND flash market shares, firmware importance, and the impact of AI, big data, and 5G on storage demand, while also highlighting major players such as Intel, Samsung, and Yangtze Memory.
