This article presents LAST, a novel Learning-at-Serving-Time approach that enables real‑time online learning for re‑ranking in industrial recommendation pipelines, eliminating feedback latency, detailing its architecture, challenges, experimental validation, and practical advantages over traditional online learning methods.
In an interview with Vivo AI engineer Liang Tianan, the article explores the challenges of post‑Q&A recommendation, the integration of large language models into recall, ranking and evaluation pipelines, and the engineering trade‑offs required to deliver high‑quality, diverse suggestions on mobile devices.
Join Baidu senior algorithm engineer Ji Zhi at the DataFun Summit 2025 to explore how generative AI transforms ad recommendation recall, covering item representation, evolving solution architectures, long‑sequence challenges, and practical insights for building efficient large‑model recommendation systems.
This article explains EasyRec's end‑to‑end recommendation system, covering its training‑inference architecture, a series of CPU/GPU and distributed optimizations, and a real‑time online‑learning pipeline that together improve throughput, latency, and model freshness.
AI infrastructuredistributed computinginference optimization
Meituan’s recommendation team introduced the MTGR framework, aligning traditional DLRM features with a unified HSTU‑based Transformer to explore scaling laws, delivering a 65‑fold FLOPs boost, 12% lower inference cost, and record gains in online CTR and order volume across its food‑delivery platform.
This announcement introduces a live session that will dissect two best‑paper award research works from WSDM 2025—one revealing how recommendation models amplify popularity bias through spectral analysis and proposing a lightweight regularizer, and the other presenting a graph disentangle causal model that integrates GNNs with structural causal models to improve causal inference on networked observational data.
The award‑winning WSDM 2025 paper reveals how recommendation models magnify popularity bias through spectral alignment of the rating matrix’s leading singular vector with item popularity, and proposes the ReSN regularization technique to mitigate this bias efficiently.
This article reviews recent advances in applying large language models to CPS advertising recommendation, outlines business requirements and core technical challenges, proposes an extensible multi‑task generative framework with explicit intent perception and multi‑objective optimization, and presents offline and online performance gains along with future research directions.
This paper investigates the "sandglass" phenomenon in residual‑quantized semantic identifiers for generative search and recommendation, analyzes its causes of path sparsity and long‑tail token distribution, and proposes heuristic and adaptive token‑removal methods that substantially improve model performance in e‑commerce scenarios.
adaptive token removalgenerative retrievalrecommendation systems
UQABench introduces the first benchmark for assessing high‑density user embeddings that serve as soft prompts in LLM‑driven recommendation, featuring a three‑stage pre‑train‑align‑evaluate pipeline, seven personalized QA tasks, and findings that transformer encoders, side‑information, simple linear adapters, and larger models markedly improve accuracy while cutting input tokens to about five percent.
Alibaba’s Taobao Group unveiled its AIGX technology suite, including the RecGPT recommendation model, the AIGB generative bidding system, and a new AI‑generated video engine, detailing open‑source benchmarks, NeurIPS workshop participation, and measurable ROI improvements for e‑commerce advertising.
This article presents a comprehensive case study on how large language models are integrated into 58.com’s real‑estate recommendation platform, detailing challenges, data adaptation, prompt and parameter optimizations, embedding generation, conversational recommendation, and future directions for multimodal and generative recommendation systems.
This article provides a step‑by‑step tutorial that walks readers through the fundamentals of traditional machine‑learning algorithms, feature‑engineering techniques, model training pipelines, evaluation metrics, and then advances to deep‑learning concepts such as MLPs, activation functions, transformers, and modern recommendation‑system models.
This comprehensive guide walks readers through the evolution of machine learning, starting with basic linear models and feature engineering, progressing through logistic regression, decision trees, and deep learning architectures like MLPs, CNNs, RNNs, and transformers, and demonstrates practical implementations with code examples and evaluation metrics.
Deep LearningEvaluation Metricsfeature engineering
This article reviews a decade of recommendation‑system research, outlines the shift from traditional listwise methods to deep‑learning models, discusses the impact of large language models and AI agents, and presents future directions such as multimodal interaction, responsible AI, cognitive modeling, and ecosystem integration.
The article presents two context‑enhanced methods—ECUP for full‑chain uplift modeling and NLGR for neighbor‑list reranking—demonstrating significant gains in coupon issuance and ranking on Meituan’s aggregation page through extensive offline and online experiments.
The article details Qunar’s hotel search ranking system evolution, covering the shift from rule‑based sorting to LambdaMart, the adoption of LambdaDNN deep models, multi‑objective MMOE architectures, multi‑scenario integration, extensive feature engineering, and experimental results demonstrating significant offline and online performance gains.
This presentation details how Alibaba Cloud's AI platform integrates big‑data pipelines, feature‑store services, and large language model capabilities to construct high‑performance search‑recommendation architectures, covering system design, training and inference optimizations, LLM‑driven use cases, and open‑source RAG tooling.
This article examines the hidden complexities of A/B experiments in short‑video recommendation feeds, explains why traditional designs produce biased results due to learning, double‑sided, and network effects, and presents practical double‑sided and community‑randomized experiment frameworks to obtain unbiased strategy evaluations.
JD Advertising accelerates its generative‑recall recommendation system by integrating NVIDIA TensorRT‑LLM, which simplifies the pipeline, injects LLM knowledge, scales to billions of parameters, and delivers over five‑fold throughput gains, one‑fifth the cost, and significant CTR improvements in both recommendation and search.
Graph Neural Networks excel in recommendation systems because they can model complex user‑item relationships, capture high‑order interactions, adapt dynamically to real‑time behavior, propagate multi‑step information, enrich contextual embeddings, alleviate data sparsity, and improve long‑tail item coverage, with practical e‑commerce case studies available for download.
This issue of the weekly AI digest explores Xiaohongshu’s NoteLLM recommendation system, compares Chinese text generation in video AI across major platforms, highlights Alibaba’s Tongyi Wanxiang breakthroughs, discusses Keras founder François Chollet’s new AGI‑focused lab, and reviews Google’s Veo 2 and Imagen‑3 advancements.
This article analyzes Xiaohongshu’s content recommendation system by reviewing two official papers, detailing the NoteLLM framework for interest discovery and the Sliding Spectrum Decomposition (SSD) method for diversified recommendations, and explaining their underlying models, loss functions, and experimental results.
Kuaishou has achieved a significant milestone with 7 papers accepted at AAAI 2025, covering diverse AI research areas including video processing, recommendation systems, and image restoration, demonstrating the company's strong research capabilities in artificial intelligence.
This article explains how large language models reshape recommendation systems by simplifying pipelines, integrating world knowledge, and leveraging scaling laws, and details the engineering steps for deploying generative recall models—including product encoding, user prompting, model training, TensorRT‑LLM optimization, and continuous performance improvements.
This article outlines Tencent's AutoML research, presenting several recent papers that introduce novel neural architecture search, feature selection, pooling, embedding size, and hyper‑parameter optimization techniques to improve the efficiency, accuracy, and scalability of large‑scale advertising recommendation systems.
The article reviews the STAR framework, a training‑free recommendation approach that leverages large language model embeddings and collaborative co‑occurrence scores to retrieve and rank items, and evaluates its performance, hyper‑parameter effects, and ablation studies against existing LLM‑based recommender methods.
Kuaishou and Tsinghua University were honored with the first‑place Qian Weichang Chinese Information Processing Science and Technology Award for their collaborative content recommendation project, which achieved international‑level innovations in explainable recommendation, bias correction, and edge intelligence, and has been applied widely in Kuaishou's platform and top academic conferences.
This article presents a KDD‑2024 accepted solution that replaces autoregressive generators with a non‑autoregressive model for video recommendation reranking, detailing the challenges, model architecture, novel loss function, extensive offline and online experiments, and practical Q&A from the conference.
The 30th SIGKDD conference in Barcelona featured 2,046 research papers with a 20% acceptance rate, and this article compiles the 59 recommendation‑system papers—covering large‑model recommenders, graph‑based methods, sequential models, fairness, privacy, advertising, debiasing, reinforcement learning and more—for researchers to explore the latest academic advances.
This article introduces the concept and capability model of strategy products, outlines the three generations of product managers, presents a simplified talent development framework, discusses practical workflow, examines six AI‑era strategic product questions, and shares 2024 observations on recommendation performance and future skill development.
This article reviews the necessity, benefits, and practical implementations of multi-task learning in e‑commerce search, detailing model selection, architecture extensions such as ESMM and ESM², and future directions for handling user behavior sequences and multi‑objective optimization.
NoteLLM-2 introduces multimodal In-Content Learning and Late Fusion to overcome visual‑modality bias in end‑to‑end fine‑tuned large representation models, delivering significant gains over baseline multimodal LLMs and traditional retrieval methods in recommendation tasks.
AI researchContrastive Learningin‑content learning
This paper presents the implementation and practice of generative retrieval based on Yuan large model in Tencent Advertising, addressing three key challenges: user intent capture, model alignment in advertising domain, and high-performance platform design under ROI constraints.
Advertising TechnologyHigh-Performance ComputingLarge Language Models
This article details the performance‑focused optimizations applied to TorchRec, PyTorch's large‑scale recommendation system library, including CUDA graph capture, multithreaded kernel launches, pinned memory copies, and input‑distribution refinements that together achieve a 2.25× speedup on MLPerf DLRM‑DCNv2 across 16 DGX H100 nodes.
This article introduces LAST, a novel online learning approach that updates recommendation models instantly at serving time, addressing real‑time learning challenges, re‑ranking complexities, and demonstrating superior offline and online performance in industrial e‑commerce scenarios.
The article details JD Advertising's technical challenges and solutions for large‑scale sparse recommendation models, describing GPU‑focused storage, compute and I/O optimizations for both training and low‑latency inference, including distributed pipelines, heterogeneous deployment, batch aggregation, multi‑stream execution, and compiler extensions.
The RECom compiler introduces a subgraph‑parallel fusion technique and symbolic shape handling to dramatically speed up GPU inference of deep recommendation models with massive embedding columns, achieving up to 6.61× lower latency and 1.91× higher throughput than TensorFlow baselines, while eliminating redundant computations.
This article outlines six essential criteria—accuracy, robustness, simplicity, explainability, generalization, and scalability—for evaluating the quality of mathematical models such as e‑commerce recommendation systems, helping readers assess whether a model is truly reliable or merely a flashy façade.
This article reviews cutting‑edge research on integrating large language models with graph‑based recommendation systems, detailing four key strategies—LLM node embeddings, deep graph‑LLM fusion, model‑driven graph data training, and text‑modal enhancements—while analyzing representation learning, InfoNCE optimization, explainable recommendations, and extensive experimental validation.
This article presents a comprehensive study on integrating multimodal image‑text representations into large‑scale e‑commerce advertising CTR models, introducing a semantic‑aware contrastive pre‑training (SCL) method and two application algorithms (SimTier and MAKE) that together achieve over 1 % GAUC improvement and significant online gains.
This article presents a comprehensive overview of recent advances in graph technology, covering GraphGPT for large language models, knowledge transfer on complex graphs, financial fraud detection, telecom network optimization, graph foundation models, Baidu's multi‑domain recommendation, high‑availability graph databases, and Kuaishou's efficient recommendation architecture.
Graph DatabasesLarge Language Modelsfinancial fraud detection
Alibaba’s advertising team introduces semantic‑aware contrastive learning to pre‑train multimodal image‑text embeddings, integrates them via SimTier and MAKE into ID‑based CTR models, achieving up to 6.9% lift in Taobao display ad click‑through rates and improving long‑tail item performance.
This article reviews the state‑of‑the‑art applications of large language models in recommendation systems, summarizing background knowledge, recent advances such as LLM4Rec, various tuning strategies, agent‑based approaches, open research problems, and future directions for generative recommendation.
Meituan’s five long papers accepted at KDD 2024 introduce a dual‑intent model for search‑recommendation, a joint auction mechanism for ads, a robust ATE estimator for heavy‑tailed metrics, a decision‑focused causal learning framework for marketing, and an efficient on‑demand order‑pooling system for real‑time courier assignments.
This article examines Meta's SUM (Scaling User Modeling) system, detailing its upstream‑downstream architecture, the SOAP online asynchronous serving platform, production optimizations, and extensive offline and online experiments that demonstrate significant gains in ad personalization performance.
The article presents Ant Group's research on integrating large language models into recommendation pipelines, covering background challenges, knowledge extraction, teacher‑model distillation, efficient deployment, experimental results, and future directions to improve accuracy and reduce bias.
Meta’s new paper introduces the Wukong model, demonstrating that expanding dense‑layer parameters and computational FLOPs in large‑scale recommendation systems follows a clear scaling law, yielding consistent performance gains across massive internal datasets, with detailed analysis of feature modules, parameter impacts, and experimental results.
On June 27, 2024, Xiaohongshu’s technical team will livestream a two‑hour session across WeChat Channels, Bilibili, Douyin and Xiaohongshu, showcasing six top‑conference papers on large‑model advances—including early‑stopping and fine‑grained self‑consistency, novel evaluation methods, negative‑sample‑assisted distillation, and LLM‑based note recommendation—followed by a Q&A and recruitment briefing.
This article discusses how Alibaba Cloud’s AI platform PAI‑Rec reduces recommendation system costs and boosts efficiency by optimizing training resources, leveraging FeatureStore, EasyRec and TorchEasyRec frameworks, detailing workflow stages, feature consistency, GPU acceleration, componentized model configuration, and practical deployment timelines.
This article presents a comprehensive overview of applying reinforcement learning to recommendation systems, covering background challenges, practical exploration, frontier research directions, multi‑agent and inverse RL approaches, evaluation methods, and future outlooks, based on a KDD‑published study and industry experience.
The DataFunSummit2024 Recommendation System Architecture Summit invites participants to explore cutting‑edge advances in large‑model recommendation, training and inference optimization, feature engineering, multi‑task modeling, and graph‑based techniques through a series of expert talks and panel discussions from leading industry and academic researchers.
This article surveys the impact of large language models on recommendation systems, outlines generative recommender architectures, discusses challenges of JD Union advertising, presents a semantic‑ID based solution with training and inference details, and reports offline and online experimental results.
This article reviews how large language models are transforming recommendation systems, covering their fundamentals, recent LLM‑enabled methods for representation, learning and generalization, challenges such as scalability, bias and privacy, and future research directions including personalized prompts and robust model integration.
This article reviews the NoteLLM paper, which leverages Llama 2 to create richer text embeddings and automatically generate tags and categories for note recommendation, addressing cold‑start issues through a multitask prompt design, generative‑contrastive learning, and collaborative supervised fine‑tuning, and demonstrates strong offline and online gains.
This article introduces the background of graph data, explains common graph modeling algorithms such as graph embedding and graph neural networks, compares their strengths, and details the evolution and large‑scale deployment of Feed graph models in Baidu's recommendation platform.
This article provides a comprehensive overview of how large language models (LLMs) are integrated into recommendation systems, detailing two main paradigms—LLM as a component and LLM as a standalone system—while discussing their impact on retrieval, ranking, prompting, and outlining future research challenges such as multimodal recommendation, hallucination mitigation, bias reduction, and agent‑based approaches.
This article reviews the evolution of JD’s recommendation advertising ranking mechanism, covering its economic auction origins, challenges of multi‑material valuation, user interest uncertainty, and multi‑item auction fairness, and describes AI‑driven solutions such as deep auction models and reinforcement‑learning‑based ListVCG.
LLMRG introduces a novel framework that leverages large language models to construct personalized reasoning graphs, integrating chain reasoning, self‑verification, divergent extension, and knowledge‑base self‑improvement, thereby enhancing recommendation accuracy, interpretability, and performance across multiple benchmark datasets without additional user or item information.
This presentation outlines the current bottlenecks of conventional recommendation pipelines and proposes a 2026 roadmap that includes retention improvement, user‑growth strategies, content‑ecosystem metrics, Pareto‑optimal multi‑objective optimization, long‑term value modeling, site‑wide spatial optimization, interactive recommendation, personalized modeling, and the integration of large‑model fusion through the OneRec framework.
Large Language ModelsMulti-Objective Optimizationinteractive recommendation
This article provides a comprehensive overview of user portrait algorithms, covering their historical development, ontology‑based traditional approaches, deep‑learning enhancements, representation‑learning techniques such as lookalike, active‑learning driven iteration, and the integration of large‑model world knowledge, while also discussing current challenges and future research directions.
This article reviews how the rapid rise of large language models, exemplified by ChatGPT, is transforming recommendation systems by addressing traditional ID‑centric limitations, introducing prompt‑based and ID‑free representations, discussing recent research advances, practical challenges, and future research directions.
This article presents a comprehensive overview of multi‑task and multi‑scenario recommendation algorithms, detailing background challenges, algorithm classifications such as TAML, CausalInt, and DFFM, their modular designs, experimental validations, and practical Q&A insights for large‑scale advertising systems.
Meta introduces a generative recommendation framework (GR) built on the Hierarchical Sequential Transduction Unit (HSTU) that unifies heterogeneous features, treats user behavior as a new modality, and leverages novel encoder and inference optimizations to achieve order‑of‑magnitude scaling in model size, training compute, and online latency while delivering 12‑18% online gains over traditional deep recommendation models.
Meta’s new Generative Recommendation (GR) paper replaces a decade‑old hierarchical feature paradigm with an ultra‑long sequence transformer that directly fuses user profiles, behaviors, and targets, offering stronger feature crossing, richer information utilization, and massive compute gains, while revealing scaling‑law effects in recommendation systems.
Vector retrieval, based on converting data into high‑dimensional vectors and measuring similarity, enables fast, accurate search across massive datasets, supporting AI tasks such as search engines, recommendation, NLP, and computer vision, and plays a crucial role in large‑model training for data selection, anomaly detection, and model optimization.
AI TrainingApproximate Nearest Neighborinformation retrieval
This collection compiles technical articles that explore diverse AI/ML applications, from deploying large language models on MacBooks and building e‑commerce recommendation engines, to leveraging the LangChain framework, creating AIGC‑driven fashion solutions, and implementing Stable Diffusion for image generation.
This article presents Kuaishou's end‑to‑end engineering solutions for handling massive, real‑time recommendation model data, covering scenario description, complex business pipelines, trillion‑parameter model storage, high‑throughput processing with Flink and NVM, and future directions for cloud‑native scalability.
This article presents Tencent TRS's industrial practice of applying meta‑learning and cross‑domain recommendation to address personalization challenges, detailing problem definitions, solution architectures, algorithmic choices such as MAML, deployment strategies, and the cost‑effective outcomes achieved across multiple scenarios.
This article explains the principles of A/B testing and the ε‑greedy multi‑armed bandit algorithm, illustrates their practical use in e‑commerce recommendation optimization, and draws broader life lessons about balancing exploration and exploitation for better personal and professional decisions.
This article presents a comprehensive study of the MIND multi‑interest recommendation model, detailing its original architecture, identified shortcomings, and proposed enhancements—including capsule initialization via max‑min and Markov methods, routing simplifications, and training adjustments—along with experimental results and business impact assessments.
Baidu’s recommendation ranking system handles billions of daily impressions and millions of users by combining discrete and cross features, bias mitigation, and long‑short sequence modeling within a multi‑stage funnel and hierarchical architecture, while planning to integrate large language models for generative, interpretable, and decision‑oriented recommendations.
This article presents Baidu's comprehensive approach to feed recommendation ranking, covering business and data background, feature engineering principles, core algorithmic strategies, system architecture design, and upcoming plans to integrate large language models for more intelligent and fair recommendations.
This article presents Ant Group's causal debiasing techniques for recommendation and marketing, covering bias background, data‑fusion based MDI model, back‑door adjustment methods, experimental results on public and industry datasets, and practical applications in advertising and e‑commerce.
This article presents a two‑stage constrained actor‑critic (TSCAC) algorithm that models short‑video recommendation as a constrained reinforcement‑learning problem, details its theoretical formulation and optimization loss, and validates its superiority through extensive offline and online experiments, followed by a multi‑task reinforcement‑learning framework (RMTL) that further improves multi‑objective recommendation performance.
The article argues that AI agents powered by large language models will become the next paradigm after search and recommendation, enabling users to actively express complex needs, reducing information asymmetry, disrupting advertising models, diminishing brand reliance, and creating an agent-centric ecosystem where data quality and security are paramount.
This article shares Baidu's experience and thinking on cross‑domain multi‑objective estimation and fusion for video recommendation, covering background, system overview, multi‑objective design and modeling, long‑term value attribution, cross‑domain network architecture, and the evolution‑strategy based fusion approach.
This article explores why recommendation engineers often struggle to deliver measurable outcomes, examining system complexity, uncertainty, delayed feedback, and personal belief, and then offers concrete principles and actionable approaches to prioritize work, align with business goals, and achieve tangible results.
This article analyzes the challenges of multi‑task and multi‑scene recommendation for large‑scale C‑end services, reviews key academic and industry solutions such as Shared‑Bottom, MMoE, PLE, ESMM, LHUC, PEPNet, MTMS and HiNet, and details ZhiZhuan’s end‑to‑end architecture that achieved over 6% click‑through and 2% conversion improvements.
This article examines recommendation system problem definitions, differences between academia and industry, offline evaluation pitfalls and data leakage issues, data construction challenges with datasets like MovieLens, and evaluates whether large language models can serve as effective solutions for modern recommendation tasks.
Large Language ModelsOffline Evaluationdata leakage
The paper introduces a real‑time controllable, multi‑objective re‑ranking framework for Taobao’s feed recommendation that combines actor‑critic reinforcement learning with hypernetworks to instantly adjust objective weights, handling diverse media and cold‑start constraints while delivering higher click‑through, diversity, and cold‑start ratios with only 20‑25 ms latency.
This article presents a comprehensive study of a scenario‑adaptive and self‑supervised multi‑scenario recommendation model (SASS) for Taobao, detailing its motivation, adaptive multi‑scenario architecture, two‑stage pre‑training and fine‑tuning, experimental validation, deployment in the recall stage, and practical challenges addressed through Q&A.
In this interview, OPPO’s senior manager Lai Hongke explains how the company tackles the challenges of sparse, cross‑scenario data in app distribution by deploying a unified modeling framework, MMOE sharing, and the oCPX capability to simultaneously cut costs, improve recommendation performance, and preserve user value across its software store and game center.
The article examines how OPPO tackles the challenges of sparse, multi‑scenario app‑distribution data by deploying a unified modeling framework, leveraging MMoe and oCPX techniques to enhance recommendation performance, reduce costs, and preserve user value across its software store and game center.
This article presents Baidu's technical experience on designing, estimating, and fusing cross-domain multi-objective models for its immersive video recommendation system, covering business background, system architecture, target design, long‑term value modeling, and evolution strategies.
This article introduces STAN, a user‑lifecycle‑aware multi‑task recommendation model proposed by Shopee that refines CTR, CVR, and stay‑time predictions by identifying and tracking user states, demonstrates offline gains on Shopee and public datasets, and reports online improvements in click‑through, dwell‑time, and order metrics.
This article summarizes a collection of newly released open‑source AI projects covering natural‑language processing, multimodal processing, intelligent agents, recommendation systems, and model training acceleration, providing brief descriptions, key capabilities, and links to their repositories.
This article presents a detailed 28‑chapter curriculum for mastering Python full‑stack development, covering Linux basics, Python fundamentals, web front‑end design with Vue, RESTful API creation with Flask, Django and Django REST Framework, big‑data processing with Hadoop, Spark and MapReduce, feature engineering, recommendation systems, and live streaming system implementation.
The HC² framework enhances multi‑scenario ad ranking by jointly applying a generalized contrastive loss on shared representations and an individual contrastive loss on scenario‑specific layers, using label‑aware positive sampling, diffusion‑noise negative sampling, and inverse‑similarity weighting, achieving consistent offline gains and up to 2.5% CVR and 3.7% GMV improvements in Alibaba’s live system.
This article presents a comprehensive study of a controllable multi‑objective re‑ranking model for Taobao's information‑flow recommendation, detailing the challenges of complex feed scenarios, three modeling paradigms (V1‑V3), an actor‑critic reinforcement learning framework with hypernet‑generated weights, and extensive online evaluation results.
The presentation introduces a two‑stage constrained actor‑critic algorithm that learns auxiliary policies for interaction signals before optimizing watch‑time under KL constraints, and a reinforcement‑learning multi‑task learning framework that models session‑level dynamics with adaptive multi‑critic weighting, both achieving significant offline and online gains in short‑video recommendation.
This article outlines the challenges of multi-task and multi-scenario modeling for large-scale C-end services, reviews key industry approaches such as Shared-Bottom, MMoE, PLE, ESMM, and LHUC, and details Zhaozhuan’s own EPNET-based solution that improved click-through and conversion rates.
At CNCC 2023, leading researchers and industry experts convened to examine how large language models can transform recommendation systems, outlining four core challenges—model integration, fluency versus intelligence, hallucination versus deception, and user understanding—while highlighting opportunities such as multimodal content, cold‑start solutions, zero‑shot ranking, instruction‑driven algorithms, and responsible, interactive recommendation pipelines.
This article presents a two‑stage constrained actor‑critic reinforcement learning algorithm for short‑video recommendation, models the problem as a constrained MDP, details the algorithm’s stages, and reports extensive offline and online experiments showing superior watch‑time and interaction metrics, followed by a multi‑task RL framework and its evaluations.
constrained optimizationmulti‑task learningrecommendation systems
This article explains the role and mindset of a strategy product manager, outlines the decision‑making framework for content recommendation platforms, compares it with related positions, and answers practical questions about value, AI impact, commercial‑consumer trade‑offs, and content creation versus consumption.
This article presents OPPO Andes Smart Cloud's intelligent growth algorithm framework for the smartphone sector, detailing industry background, data and model architecture, four real-world application cases—including AIGC content generation, multimodal recommendation, causal inference, and precise advertising—and summarizing key insights from a technical Q&A session.
This article examines how large language models fit into recommendation systems by discussing problem definitions, offline evaluation pitfalls such as data leakage, dataset construction issues exemplified by MovieLens, and the practical limits of using LLMs as a universal solution.
This article outlines the three key career stages for algorithm engineers, offers practical planning advice through vision, self‑evaluation and action, and discusses industry trends, skill‑building paths, mindset, and work‑life balance to help engineers navigate a volatile tech landscape.