arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

ACM SIGKDD Conference on Knowledge Discovery and Data Mining · 会议 · Data Mining

共收录 2626 篇
2611.17899 2026-11-16 cs.CV cs.AI q-bio.QM 新提交

More Genes, Faster Inference: Efficient Large-Panel Spatial Transcriptomics Prediction from Histology

更多基因,更快推理:基于组织学的高效大面板空间转录组预测

Songming Zhang, Yuxiao Luo, Shifu Luo, Siran Chen, Shuquan Su, Jinyan Li

机构 * School of Computer Science, Nanjing University(南京大学计算机科学学院) ; Faculty of Computer Engineering and Artificial Intelligence, Shenzhen University of Advanced Technology(深圳先进技术学院计算机工程与人工智能学院) ; Faculty of Health Sciences, University of Macau(澳门大学健康科学学院) ; Tencent(腾讯)

AI总结 研究人员提出GeneFlow框架,将流匹配生成建模与结构化多模态融合结合,实现高效大面板空间转录组预测,在1000基因预测中获最高56.8%提升,计算成本大幅降低,可用于临床及生物学发现。

Comments Accepted at KDD 2027

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.17878 2026-11-16 cs.CV cs.AI 新提交

DySENSE: Dynamic Spatiotemporal ENergy Synthesis and Emulation with Satellite Imagery

DySENSE:基于卫星影像的动态时空能源合成与仿真

Kailai Sun, Zijie Gao, Mingyi He, Can Rong, Alok Prakash, Baoshen Guo, Shenhao Wang, Jinhua Zhao

机构 * SMART ; Massachusetts Institute of Technology(麻省理工学院) ; National University of Singapore(新加坡国立大学) ; University of Florida(佛罗里达大学)

AI总结 本文提出DySENSE多模态GenAI框架,解耦城市静态形态与动态能耗,用不足25%标注数据生成高保真合成能源数据,提升下游预测性能,为城市能源相关研究提供有效基础。

Comments Accpted by KDD 2027

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.25096 2026-11-13 cs.AI cs.IR 版本更新

Toward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data Sources

面向企业零售工作流的智能体检索增强生成:针对异构数据源

Jeonghwan Choi, Jibin Hwang, Hayoon Park, Hyeonjae Cheon, Hwanjun Song

AI总结 本研究针对企业零售工作流的异构数据源串联挑战,构建RetailBridge基准,提出RetailRAG框架,评估发现多源串联是瓶颈,为零售智能体评估提供参考。

Comments Accepted to KDD 2027 Datasets and Benchmarks Track (Cycle 1)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2603.02184 2026-11-13 cs.LG cs.AI 版本更新

MAC: A Conversion Rate Prediction Benchmark Featuring Labels Under Multiple Attribution Mechanisms

MAC: 一个包含多种归因机制标签的转化率预测基准

Jinqi Wu, Sishuo Chen, Zhangming Chan, Yong Bai, Lei Zhang, Sheng Chen, Chenghuan Hou, Xiang-Rong Sheng, Han Zhu, Jian Xu, Bo Zheng, Chaoyou Fu

机构 * .1em blue n Nanjing University, Nanjing, China .1em purple a Taobao \& Tmall Group of Alibaba, Beijing, China ; .1em blue n Nanjing University, Nanjing, China ; .1em purple a Taobao \& Tmall Group of Alibaba, Beijing, China

AI总结 本文提出MAC基准和PyMAL库,通过多归因机制标签提升转化率预测性能,提出MoAE方法在多归因学习中取得显著效果。

Comments 11 pages, 5 figures. Updated to the camera-ready version published in the proceedings of KDD 2026. Code and data available at https://github.com/alimama-tech/PyMAL

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '26), pp. 10009-10019, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2406.04428 2026-11-13 cs.CL cs.AI 版本更新

MoralBench: Moral Evaluation of LLMs

MoralBench:大语言模型的道德评估

Jianchao Ji, Yutong Chen, Mingyu Jin, Wujiang Xu, Wenyue Hua, Yongfeng Zhang

机构 * Rutgers The State University of New Jersey(新泽西州立大学拉特格斯分校) ; University of Chicago(芝加哥大学)

AI总结 本文提出MoralBench基准及数据集,用于系统评估和比较大语言模型的道德推理能力,并通过多模型实验揭示其道德表现差异与局限性。

Comments Accepted to ACM SIGKDD Explorations Volume 27 Issue 1

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.14493 2026-11-11 cs.AI cs.CL cs.LG cs.SE 新提交

Comparative Analysis of Agent Evaluation Frameworks: Stability, Detection, and Discovery in Enterprise Analytics Agents

智能体评估框架的对比分析:企业分析智能体中的稳定性、检测与发现

Howard Zhang, Chuck Arvin, Logan Brown, Mihir Bhatia, Neelu Choudhary, Lily Miao

机构 * Amazon(亚马逊)

AI总结 该研究对比三种评估框架在企业分析智能体中的表现,发现基于轨迹的框架性能最优,功能与安全评估互补,建议采用双框架策略。

Comments Accepted to KDD 2026, Jeju, Korea. 9 pages, 3 figures

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.12725 2026-11-10 cs.LG cs.IR 新提交

Scaling Taxonomy-Free Image Labeling with Generative AI: A Multimodal

用生成式AI扩展无分类学的图像标注:一种多模态方法

Zhenjie Zhang, Sainan Chen, Kshiteesh Hegde, Mengze, Li, Pouya Rezazadeh Kalehbasti, Munachim Ezema, Karina Sobhani

机构 * Pinterest, Inc.(Pinterest公司)

AI总结 本研究提出基于集群的生成式标注(GLOC),用于解决传统图像标注方法的局限,在Pinterest的混合搜索中验证其性能优于过往方法,实现了高可扩展、低成本的多模态图像洞察服务。

Comments 5 pages, 3 figures, 4th CJ Workshop at KDD 2025 (https://sites.google.com/view/kddcjworkshop2025/accepted-papers), also related: https://arxiv-org.atlassian.net/servicedesk/customer/portal/2/MOD-93955

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.12509 2026-11-10 cs.LG cond-mat.mtrl-sci 新提交

UniHam: A Large-Scale SOC-Complete Dataset and Benchmark for Hamiltonian Learning in Materials

UniHam:面向材料哈密顿学习的大规模SOC完备数据集与基准

Yuewen Huang, Pin Chen, Yutong Lu

机构 * Sun Yat-Sen University(中山大学)

AI总结 该研究构建了含10万+个完整SOC哈密顿量的UniHam数据集与基准,测试了现有模型并引入多维度评估指标,为量子材料基础模型开发提供支撑。

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Vol. 2 (2026), pp. 9486-9497

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2606.15874 2026-11-10 cs.AI cs.SE 版本更新

LLM-as-Code: Agentic Programming for Agent Harness

LLM即代码:面向Agent框架的编程范式

Junjia Qi, Zichuan Fu, Jingtong Gao, Wenlin Zhang, Hanyu Yan, Xian Wu, Xiangyu Zhao

机构 * City University of Hong Kong(香港城市大学) ; Tencent Jarvis Lab(腾讯贾维斯实验室)

AI总结 针对LLM作为编排器导致控制流幻觉和不可靠执行的问题,提出Agentic Programming范式,由程序控制所有流程,LLM仅作为代码组件在需要推理或生成时被调用,显著提升长序列操作的稳定性。

Comments Accepted at the KDD 2026 Workshop on Agentic Software Engineering (AgenticSE)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2506.10912 2026-11-05 cs.AI cs.CL 版本更新

Breaking Bad Molecules: Are MLLMs Ready for Structure-Level Molecular Detoxification?

Breaking Bad Molecules: MLLMs 是否准备好进行结构级分子解毒?

Fei Lin, Ziyang Gong, Cong Wang, Tengchao Zhang, Yonglin Tian, Yining Jiang, Ji Dai, Chao Guo, Xiaotong Yu, Xue Yang, Gen Luo, Fei-Yue Wang

机构 * Department of Engineering Science, Macau University of Science and Technology, Macau, China(澳门科学技术大学工程科学系) ; School of Computer Science, Shanghai Jiao Tong University, Shanghai, China(上海交通大学计算机科学学院) ; Institute of Automation, Chinese Academy of Sciences, Beijing, China(中国科学院自动化研究所) ; School of Pharmacy, Macau University of Science and Technology, Macau, China(澳门科学技术大学药学院) ; Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, China(宁波大学电气与计算机科学学院) ; State Key Laboratory of Biopharmaceutical Preparation and Delivery, Institute of Process Engineering, Chinese Academy of Sciences, Beijing, China(中国科学院生物制药制备与递送国家重点实验室) ; School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Shanghai, China(上海交通大学自动化与智能感知学院) ; Shanghai Artificial Intelligence Laboratory, Shanghai, China(上海人工智能实验室)

AI总结 本文提出 ToxiMol 基准任务,利用多模态大语言模型进行分子毒性修复,并构建数据集、提示流程和自动评估框架 ToxiEval,实验表明当前模型虽面临挑战但展现出毒性理解与结构编辑的潜力。

Comments Accepted by KDD 2026 (AI for Sciences Track, Oral)

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '26), Jeju Island, Republic of Korea, pp. 11399-11410, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
1903.07745 2026-11-03 stat.ML cs.LG 版本更新

Attention-Based Multiple Instance Regression for Remote Sensing

基于注意力机制的多实例回归在遥感中的应用

Thomas Uriot

机构 * European Space Agency(欧洲空间局)

AI总结 针对遥感中多实例回归问题,提出不依赖核心实例假设的注意力机制方法,在5个真实数据集上3个优于现有最先进模型,添加特征矩可提升准确率但收益递减。

Comments KDD 2019, FEED Workshop

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.41368 2026-11-02 cs.AI 新提交

RECITE: Revising Eligibility Criteria Incorporating Textual Evidence

RECITE:结合文本证据修订合格标准

Russell Ro, Reza Abbasi-Asl

机构 * University of California, San Francisco(加利福尼亚大学旧金山分校) ; University of California, Berkeley(加利福尼亚大学伯克利分校)

AI总结 针对临床试验合格标准过严导致入组失败的问题,提出RECITE框架,构建相关数据集并测试15种LLM,发现Gemma 2 9B、GPT-4o表现良好,还识别出44篇论文-试验匹配项,证实临床知识可转化为试验设计建议。

Comments 18 pages, 4 figures, 4 tables. Extended version of a paper in KDD 2026 (32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining). Code and data: https://doi.org/10.5281/zenodo.20487342

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2507.23267 2026-11-02 cs.IR 版本更新

Your Spending Needs Attention: Modeling Financial Habits with Transformers

你的支出需要关注:用Transformer建模金融习惯

D. T. Braithwaite, Misael Cavalcanti, R. Austin McEver, Hiroto Udagawa, Daniel Silva, Rohan Ramanath, Felipe Meneses, Arissa Yoshida, Evan Wingert, Matheus Ramos, Brian Zanfelice, Aman Gupta

机构 * Nubank

AI总结 该研究针对金融数据利用难题,提出基于Transformer的nuFormer模型,结合自监督学习处理交易数据,在Nubank大规模推荐任务上仅通过增强表示学习实现性能提升。

Comments Presented at 9th Workshop on Machine Learning in Finance at KDD 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2403.15524 2026-10-30 cs.GT cs.LG 版本更新

PPA-Game: Characterizing and Learning Competitive Dynamics Among Online Content Creators

PPA-Game:刻画与学习在线内容创作者之间的竞争动态

Renzhe Xu, Haotian Wang, Xingxuan Zhang, Bo Li, Peng Cui

机构 * MoE Key Laboratory of Interdisciplinary Research of Computation and Economics, Institute for Theoretical Computer Science, Shanghai University of Finance and Economics(交叉学科计算与经济学联合实验室,理论计算机科学研究院,上海财经大学) ; College of Computer Science and Technology, National University of Defense Technology(计算机科学与技术学院,国防科技大学) ; Department of Computer Science and Technology, Tsinghua University(计算机科学与技术系,清华大学) ; School of Economics and Management, Tsinghua University(经济管理学院,清华大学)

AI总结 本文提出比例收益分配博弈(PPA-Game)以刻画在线内容创作者竞争注意力的动态,通过博弈论分析纯纳什均衡存在性,并设计多臂老虎机在线学习算法,实现遗憾界O(log^{1+η}T)。

Comments KDD 2025

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.37316 2026-10-28 cs.AI 新提交

ReForest: Interpretability Meets Reasoning in Isolation Forest

ReForest:可解释性与推理在孤立森林中的结合

Juyoung Seo, Geonwoo Bang, Dongho Kim, Moohong Min

机构 * Sungkyunkwan University(成均馆大学)

AI总结 ReForest框架结合孤立森林的统计效率与LLMs的推理能力,通过将孤立森林决策路径转为自然语言解释,经动态集成优化异常检测,在医疗、金融等数据集上召回率平均提升21%、F1分数提高10%,兼具高性能与可解释性。

Comments 12 pages, 2 figures. Published in PAKDD 2026

Journal ref Advances in Knowledge Discovery and Data Mining (PAKDD 2026), Lecture Notes in Computer Science, vol. 16600, pp. 628-639, Springer (2026)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2507.15042 2026-10-28 cs.AI cs.IR

DeRAG: Black-box Adversarial Attacks on Multiple Retrieval-Augmented Generation Applications via Prompt Injection

DeRAG:通过提示注入对多个检索增强生成应用的黑盒对抗攻击

Jerry Wang, Fang Yu

机构 * Department of Management Information Systems, National ChengChi University(管理信息系,国立中正大学)

AI总结 本文提出DeRAG方法,利用差分进化黑盒优化对抗性提示后缀,在仅使用不超过5个token的情况下提升错误文档检索排名,实现对多个RAG应用的攻击,并验证其后缀可读性及对BERT检测器的规避能力。

Comments Accepted by KDD Workshop on Prompt Optimization 2025

Journal ref Proceedings of the First International KDD Workshop on Prompt Optimization, KDD 2025, Toronto, Canada, 2025

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.32189 2026-10-26 cs.LG 新提交

EVOLVE: Uncertainty-Aware Multivariate Time Series Anomaly Prediction with Adaptive Prototype Memories

EVOLVE:基于自适应原型记忆的不确定性感知多变量时间序列异常预测

Jianxin Jin, Kai Zhao, Shiyan Hu, Yang Shu, Peng Chen, Meng Wang, Zhongwen Rao, Yijie Li, Jing Wang, Bin Yang, Chenjuan Guo

机构 * East China Normal University(华东师范大学) ; Aalborg University(奥尔堡大学) ; Huawei(华为) ; Huawei Technologies Co., Ltd(华为技术有限公司) ; Huawei Cloud(华为云)

AI总结 本文提出EVOLVE框架,针对现有无监督多变量时间序列异常预测方法未考虑异常演化的变量与上下文依赖性及不确定性的问题,通过自适应原型记忆与不确定性感知推理实现更优的异常预测性能。

Comments 14 pages. An earlier version of this work was submitted to KDD 2026 on February 9, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.22415 2026-10-19 eess.IV cs.LG 新提交

Deep Learning for Forest Plantation Mapping in Godavari Districts of Andhra Pradesh, India

印度安得拉戈达瓦里地区人工林制图的深度学习方法

Snehal More, Anuj Karpatne, Valerie A. Thomas, Randolph H. Wynne

机构 * Virginia Tech(弗吉尼亚理工大学)

AI总结 本研究针对印度安得拉邦戈达瓦里地区小型人工林制图难题,采用深度学习方法,结合Harmonized Landsat Sentinel-2 S10数据与实地标注训练集,对比随机森林分类器,分析不同土地覆盖及季节的分类结果差异。

Journal ref KDD 2019 Earth Day Workshop

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2507.04651 2026-10-14 cs.IR

FindRec: Stein-Guided Entropic Flow for Multi-Modal Sequential Recommendation

FindRec:面向多模态序列推荐的 Stein 引导熵流方法

Maolin Wang, Yutian Xiao, Binhao Wang, Sheng Zhang, Shanshan Ye, Wanyu Wang, Hongzhi Yin, Ruocheng Guo, Zenglin Xu

机构 * City University of Hong Kong(香港城市大学) ; Beihang University(北京航空航天大学) ; University of Technology Sydney(悉尼科技大学) ; The University of Queensland(昆士兰大学) ; Fudan University(复旦大学)

AI总结 FindRec 提出一种多模态序列推荐框架,通过 Stein 核信息协调模块和跨模态专家路由机制解决分布差异与噪声干扰问题,在长序列和噪声场景下显著提升推荐性能与可解释性。

Comments Accepted by KDD 2025

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2602.22647 2026-10-08 cs.IR cs.CL cs.LG 版本更新

Vectorizing the Trie: Efficient Constrained Decoding for LLM-based Generative Retrieval on Accelerators

将尝试矢量化:为基于LLM的生成检索在加速器上实现高效的受限解码

Zhengyang Su, Isay Katsman, Yueqi Wang, Ruining He, Lukasz Heldt, Raghunandan Keshavan, Shao-Chuan Wang, Xinyang Yi, Mingyan Gao, Onkar Dalal, Lichan Hong, Ed Chi, Ningren Han

机构 * Yale University(耶鲁大学) ; Google DeepMind(谷歌DeepMind)

AI总结 STATIC通过矢量化技术实现高效受限解码,显著提升基于LLM的生成检索性能。

Comments KDD 2026 camera-ready

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '26), 2026, pp. 8009-8020

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.07848 2026-10-07 cs.CL cs.AI 新提交

Dynamic Positional Attention Modulation for Parameter-Efficient Fine-Tuning of Large Language Models

动态位置注意力调制用于大型语言模型的参数高效微调

Dayan Pan, Jingyuan Wang, Xie Yu

机构 * Beihang University(北京航空航天大学)

AI总结 针对现有PEFT方法均匀静态适应的问题,提出DyPAM,通过维度、头、层调制位置注意力,在数学和常识推理基准上优于强基线。

Comments Accepted by KDD 2026

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '26), 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2602.23717 2026-10-06 cs.IR

Recommending Search Filters To Improve Conversions At Airbnb

为提升Airbnb的转化率推荐搜索过滤器

Hao Li, Kedar Bellare, Siyu Yang, Sherry Chen, Liwei He, Stephanie Moyerman, Sanjeev Katariya

机构 * Airbnb, Inc.(爱彼迎公司)

AI总结 本文提出利用机器学习推荐搜索过滤器以提升Airbnb的预订转化率,通过解决冷启动等挑战,成功部署系统并验证了其有效性。

Comments Accepted at the KDD 2026 Workshop on Two-sided Marketplace Optimization: Search, Discovery, Matching, Pricing & Growth (TSMO)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.38249 2026-10-02 cs.CR 版本更新

SURE: Framework for Safety to Construct Trustworthy AI

SURE:构建可信人工智能的安全框架

Soeun Han, Jisoo Lee, Jeongyong Shim, Eunkyeong Lee, Eunmi Kim

机构 * Korea Telecom(KT)(韩国电信)

AI总结 提出SURE框架,通过分类对抗提示、定义理想响应模板和绝对安全评分,实现可定制且可验证的人工智能安全对齐。

Comments 14 pages, 2 figures, 8 tables. Accepted to the 4th Workshop on Ethical Artificial Intelligence: Methods and Applications (EAI) at KDD 2025

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2607.18270 2026-10-02 cs.AI cs.CL 交叉投稿

Trajectory-Aware Clinical Risk Prediction via Severity-Grounded Knowledge Graphs and Retrieval-Augmented Generation

通过基于严重程度的知识图谱和检索增强生成进行轨迹感知临床风险预测

Kyunghoon Jeon, Youmin Ko, Woohwan Jung, Hyunjoon Kim

机构 * Hanyang University(汉阳大学) ; Korea University(韩国大学)

AI总结 针对利用EHRs预测临床风险的挑战,提出TRACER框架,通过构建含严重程度信息的知识图谱、检索患者进展路径、提取临床事件及用相似病例增强患者背景来预测风险,在相关数据集实验中相比基线有显著提升。

Comments Accepted to KDD 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.39215 2026-10-01 cs.LG cs.AI 新提交

In a Streaming World, Should You Stand Still? A Comprehensive Benchmark of Anomaly Detection in Streams

在流式世界中,你是否应该静止不动?流式异常检测的综合基准

Magali Parrino, Antoine Ajenjo, Emmanuel Remy, Pierre Stephan, Pierre Senellart, Paul Boniol

机构 * EDF R&D(法国电力集团研发部) ; ENS, PSL University, CNRS, Inria(巴黎高等师范学院,巴黎文理研究大学,法国国家科学研究中心,法国国家信息与自动化研究所) ; DI ENS, ENS, PSL University, CNRS, Inria(巴黎高等师范学院计算机科学系,巴黎高等师范学院,巴黎文理研究大学,法国国家科学研究中心,法国国家信息与自动化研究所)

AI总结 本研究通过大规模统一基准比较流式与静态时间序列异常检测方法,发现静态方法在多数流式场景中显著更优,并提出了分布漂移数据集TSB-drift,揭示现有流式方法设计与实际需求间的关键差距。

Journal ref Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.33073 2026-09-29 cs.LG cs.AI cs.CL cs.IR 新提交

Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems

生成模型增强推荐系统相关的算法危害

Christine Herlihy, Xumei Xi, Shloka Desai, Kevin Bannerman Hutchful, Pedro Silva

AI总结 本研究扩展了算法危害分类法,分析生成模型增强推荐系统可能引发的新型及内生危害,并提供因果分析以支持检测与缓解。

Comments Presented at the KDD 2025 Workshop on Online and Adaptive Recommender Systems (OARS), August 3, 2025, Toronto, Ontario, Canada

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.30396 2026-09-28 stat.ME cs.LG stat.AP stat.ML 新提交

An End-to-End Pipeline for Causal ML with Continuous Treatments: An Application to Financial Decision Making

具有连续治疗变量的因果机器学习端到端流水线:金融决策应用

Javier Moral Hernández, Clara Higuera-Cabañes, Álvaro Ibraín

机构 * BBVA(毕尔巴鄂比斯开银行)

AI总结 本文提出一个六步骤端到端因果机器学习流水线,处理连续治疗变量,包含正性违反检测、两阶段降维等方法,在金融债务催收合成数据上验证,比标准方法更高效且偏差更小。

Comments Oral presentation at the 3rd Workshop on Causal Inference and Machine Learning in Practice, KDD 2025, Toronto. Code: https://github.com/javiermoralh/causal-pipeline

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2605.25749 2026-09-25 cs.IR cs.AI cs.LG 版本更新

DeGRe: Dense-supervised Generative Reranking for Recommendation

DeGRe: 密集监督的生成式重排序用于推荐

Chaotian Song, Jingyao Zhang, Chenghao Chen, Zisen Sang, Dehai Zhao, Guodong Cao, Boxi Wu, Deng Cai, Jia Jia

机构 * College of Software, Zhejiang University Hangzhou China ; Rajax Network Technology, Taobao Shangou of Alibaba Hangzhou China ; Rajax Network Technology, Taobao Shangou of Alibaba Beijing China ; State Key Lab of CAD\&CG, Zhejiang University Hangzhou China ; Rajax Network Technology, Taobao Shangou of Alibaba Shanghai China ; College of Software, Zhejiang University ; Rajax Network Technology, Taobao Shangou of Alibaba ; State Key Lab of CAD\&CG, Zhejiang University

AI总结 提出DeGRe框架,通过离线探索中的密集监督信号(Lookahead Evaluator)指导在线生成器(Online Generator)进行单步贪婪解码,解决重排序中的启发式标签偏差和信用分配问题。

Comments Accepted to KDD 2026 ADS Track (Oral). Best Paper Award Honorable Mention

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.23102 2026-09-22 cs.LG 新提交

When Does Adversarial Refinement Help? A Negative Result and Open Problem in Adapting R3GAN to Time Series Imputation

对抗性细化何时有帮助?将R3GAN适配到时间序列插补的负结果与开放问题

Yufeng He

机构 * The University of Hong Kong(香港大学)

AI总结 本研究将R3GAN适配到时间序列插补,发现对抗性细化在多数配置下无效或有害,并指出判别器无法提供有效细化梯度的原因是一个开放问题。

Comments 4 pages, 1 figure, 2 tables. Accepted at the 12th Workshop on Mining and Learning from Time Series (MiLeTS 2026), held with KDD 2026. This arXiv version incorporates a post-workshop reproducibility audit of the saved runs. Code: https://github.com/he-yufeng/adversarial-refinement-imputation

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2609.21281 2026-09-21 cs.IR cs.DC cs.LG cs.PF 新提交

Hybrid GPU-CPU Retrieval for Personalized Search at Ultra-Large Scale

超大规模个性化搜索的GPU-CPU混合检索

Hao Fu, Jichao Sun, Baiting Zhu, Qiaoling Liu, Yan Shi, Cheng Lu, Liu Liu, Yubo Wang, Xin Yao, Xiangyu Niu, Xu Dong, Wenhan Lyu, Chiyao Shen, Yinjie Huang, Minglei Chen, Shuai Ding, Li Fan, Xiao Kong

机构 * Meta Platforms, Inc.(元平台公司)

AI总结 针对万亿级个性化搜索的规模悖论,提出GPU-CPU混合协同检索系统,GPU处理深度交互预排序,CPU覆盖更广库存,经A/B测试验证提升相关性与参与度。

Comments 10 pages, 5 figures, 9 tables. ACM sigconf format; submitted to the KDD 2027 Applied Data Science Track

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
↑