Annual Review of Statistics and Its Application最新文献

Reinforcement Learning from Human Feedback: A Statistical Perspective 基于人类反馈的强化学习：统计学视角

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-11-02 DOI: 10.1146/annurev-statistics-042925-123430

Pangpang Liu,Chengchun Shi,Will Wei Sun

{"title":"Reinforcement Learning from Human Feedback: A Statistical Perspective","authors":"Pangpang Liu,Chengchun Shi,Will Wei Sun","doi":"10.1146/annurev-statistics-042925-123430","DOIUrl":"https://doi.org/10.1146/annurev-statistics-042925-123430","url":null,"abstract":"<div><p>Reinforcement learning from human feedback (RLHF) has emerged as a central framework for aligning large language models (LLMs) with human preferences. Despite its practical success, RLHF raises fundamental statistical questions because it relies on noisy, subjective, and heterogeneous feedback to learn reward models and optimize policies. This review provides a statistical perspective on RLHF, focusing primarily on LLM alignment. We introduce the main components of RLHF, including supervised fine-tuning, reward modeling, and policy optimization, and relate them to familiar statistical concepts such as the Bradley–Terry–Luce model, latent utility estimation, active learning, experimental design, and uncertainty quantification. We review methods for learning reward functions from pairwise preference data and for optimizing policies through both two-stage RLHF pipelines and emerging one-stage approaches such as direct preference optimization. We further discuss recent extensions including reinforcement learning from artificial intelligence feedback, inference-time algorithms, and reinforcement learning from verifiable rewards, as well as benchmark datasets, evaluation protocols, and open-source frameworks that support RLHF research. We conclude by highlighting open challenges in RLHF.</p></div>","PeriodicalId":48855,"journal":{"name":"Annual Review of Statistics and Its Application","volume":"13 1","pages":""},"PeriodicalIF":7.9,"publicationDate":"2026-11-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149360078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The Noisy Work of Uncertainty Visualization 不确定性可视化的噪声工作

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-10-27 DOI: 10.1146/annurev-statistics-042825-110607

Harriet Mason,Dianne Cook,Sarah Goodwin,Emi Tanaka,Susan VanderPlas

引用次数: 0

A Statistical Perspective on Knowledge Distillation: Foundations, Classical Methods, and Large Language Model Extensions 知识蒸馏的统计视角：基础、经典方法和大型语言模型扩展

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-09-17 DOI: 10.1146/annurev-statistics-043025-102547

Luyang Fang, Haoran Lu, Jiazhang Cai, Tao Wang, Huimin Cheng, Wenxuan Zhong, Ping Ma

{"title":"A Statistical Perspective on Knowledge Distillation: Foundations, Classical Methods, and Large Language Model Extensions","authors":"Luyang Fang, Haoran Lu, Jiazhang Cai, Tao Wang, Huimin Cheng, Wenxuan Zhong, Ping Ma","doi":"10.1146/annurev-statistics-043025-102547","DOIUrl":"https://doi.org/10.1146/annurev-statistics-043025-102547","url":null,"abstract":"Knowledge distillation (KD) has emerged as a vital paradigm for transferring the capabilities of high-capacity models to efficient student counterparts, addressing critical challenges in computational cost, deployment constraints, and privacy-sensitive settings. Although KD is widely used in practice, it is often viewed primarily as an engineering technique, with a unified statistical perspective remaining less developed. This review bridges that gap by presenting a unified Bayesian formulation that formulates teacher predictions as prior information. This provides a principled interpretation of how teacher information is incorporated into student learning and establishes a rigorous connection to uncertainty quantification. We demonstrate how this foundational lens reconciles classical distillation with modern extensions in generative and foundation-model systems, showing that contemporary developments remain rooted in these same statistical principles. By synthesizing theory with emerging methodologies and diverse applications, this review provides a conceptual road map and identifies critical open problems for the future of the field.","PeriodicalId":48855,"journal":{"name":"Annual Review of Statistics and Its Application","volume":"29 1","pages":""},"PeriodicalIF":7.9,"publicationDate":"2026-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149019964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Statistics for Climate Science: Differentiation and Integration 气候科学统计：分化与整合

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-09-08 DOI: 10.1146/annurev-statistics-042424-022358

Bo Li,Trevor Harris

{"title":"Statistics for Climate Science: Differentiation and Integration","authors":"Bo Li,Trevor Harris","doi":"10.1146/annurev-statistics-042424-022358","DOIUrl":"https://doi.org/10.1146/annurev-statistics-042424-022358","url":null,"abstract":"Climate science increasingly relies on statistical tools to evaluate and synthesize diverse datasets, including observations, reanalyses, paleoclimate proxies, and model simulations. Climate data differentiation, which quantifies discrepancies between datasets, and climate integration, which combines information across them, play central roles in these efforts. This article reviews modern statistical approaches for climate data differentiation, including methods for quantifying discrepancies in means, dependence structures, distributions, extremes, spatial signals, and shapes. It also reviews approaches for climate data integration that combine information across models and observations through ensemble methods, bias correction, hierarchical modeling, and machine learning. We highlight how these methods address challenges such as high dimensionality, spatial dependence, limited observations, and structural model uncertainty. Together, these tools provide complementary perspectives for understanding differences among climate data sources and for synthesizing information across them, ultimately improving estimates of past, present, and future climate conditions.","PeriodicalId":48855,"journal":{"name":"Annual Review of Statistics and Its Application","volume":"60 1","pages":""},"PeriodicalIF":7.9,"publicationDate":"2026-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148932636","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Statistical Methods in Generative Artificial Intelligence 生成式人工智能中的统计方法

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-08-14 DOI: 10.1146/annurev-statistics-042825-121454

Edgar Dobriban

引用次数: 0

Recent Advances in Conditional Extreme Quantile Analysis 条件极值分位数分析的新进展

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-03-09 DOI: 10.1146/annurev-statistics-042324-014139

Yiwei Tang, Junyu Chen, Deyuan Li, Huixia Judy Wang

引用次数: 0

Statistical Modeling of the Ocean Environment 海洋环境的统计模型

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-03-09 DOI: 10.1146/annurev-statistics-042424-115755

Erik Vanem, Sheng Dong, Guillaume de Hauteclocque, Thomas Berge Johannessen, Tingyao Zhu, Jasna Prpic-Orsic, Sanne van Essen, Kevin Ewans, Ed Mackay, Philip Jonathan

{"title":"Statistical Modeling of the Ocean Environment","authors":"Erik Vanem, Sheng Dong, Guillaume de Hauteclocque, Thomas Berge Johannessen, Tingyao Zhu, Jasna Prpic-Orsic, Sanne van Essen, Kevin Ewans, Ed Mackay, Philip Jonathan","doi":"10.1146/annurev-statistics-042424-115755","DOIUrl":"https://doi.org/10.1146/annurev-statistics-042424-115755","url":null,"abstract":"Statistical modeling of the ocean environment is important for many practical applications in science and engineering. Probabilistic descriptions of the ocean environment are important input for structural design and risk assessment of marine structures, including ships, offshore and coastal structures, and aquaculture installations. They are also essential for the safe operation of ships and other structures at sea. Additionally, they are critical for planning and decision-making in the exploitation of marine renewable energy sources such as waves, tides, and offshore wind. This article presents a review of recent developments with regard to statistical modeling of the ocean environment, with a particular focus on ocean waves. Such developments are driven by an increasing volume of available data, increasing computational capabilities, and demand from the industry for more accurate and uncertainty-aware descriptions of relevant oceanic variables. Hence, statistical modeling of the ocean environment remains an active area of research, with significant developments in various directions. These are reviewed in this article.","PeriodicalId":48855,"journal":{"name":"Annual Review of Statistics and Its Application","volume":"60 1","pages":""},"PeriodicalIF":7.9,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Theory of Random Forests 随机森林理论

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2026-03-09 DOI: 10.1146/annurev-statistics-112723-034707

Erwan Scornet, Giles Hooker

引用次数: 0

Deep Spatiotemporal Point Processes: Advances and New Directions 深度时空点过程：进展与新方向

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2025-12-22 DOI: 10.1146/annurev-statistics-042324-040052

Xiuyuan Cheng, Zheng Dong, Yao Xie

{"title":"Deep Spatiotemporal Point Processes: Advances and New Directions","authors":"Xiuyuan Cheng, Zheng Dong, Yao Xie","doi":"10.1146/annurev-statistics-042324-040052","DOIUrl":"https://doi.org/10.1146/annurev-statistics-042324-040052","url":null,"abstract":"Spatiotemporal point processes model discrete events distributed in space and time, with applications in criminology, seismology, epidemiology, and social networks. Classical models rely on parametric kernels, limiting their ability to capture heterogeneous, nonstationary dynamics. Recent advances integrate deep neural architectures, either by modeling the conditional intensity directly or by learning flexible, data-driven influence kernels. This article reviews the deep influence kernel approach, which balances statistical interpretability by retaining explicit kernels to capture event propagation, with expressive power from neural architectures. We outline key components, including functional basis decomposition, graph neural networks for encoding spatial or network structures, and both likelihood-based and likelihood-free estimation methods, while addressing scalability for large data. We also highlight theoretical results on kernel identifiability. Applications in crime analysis, earthquake aftershock prediction, and sepsis modeling demonstrate the framework's effectiveness. We conclude with promising directions for developing explainable and scalable deep kernel point processes.","PeriodicalId":48855,"journal":{"name":"Annual Review of Statistics and Its Application","volume":"84 1","pages":""},"PeriodicalIF":7.9,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145908310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The Why and How of Convex Clustering 凸聚类的原因和方法

IF 7.9 1区数学

Annual Review of Statistics and Its Application Pub Date : 2025-12-03 DOI: 10.1146/annurev-statistics-112723-034107

Eric C. Chi, Aaron J. Molstad, Zheming Gao, Jocelyn T. Chi

引用次数: 0