arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

大厂专区

Broadcom(博通)

共收录 21
2608.18147 2026-08-20 cs.LG cs.AI cs.DS cs.IT math.IT 新提交

Entropy-Constrained Adaptive Stochastic Quantization

熵约束自适应随机量化

Ran Ben Basat, Yaniv Ben-Itzhak, Michael Mitzenmacher, Shay Vargaftik

机构 * University College London(伦敦大学学院) VMware Research by Broadcom(博通旗下VMware研究院) Harvard University(哈佛大学)

AI总结 本文针对自适应随机量化未考虑后续熵编码的精度损失问题,提出ECASQ方法,给出最优与GPU友好近似动态规划方案,实验表明其精度接近最优且速度优势显著。

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.06046 2026-08-07 cs.NI cs.DC cs.LG 新提交

ML-for-ML

面向机器学习的机器学习

Yutong Zhao, Noga H. Rotman, Gianni Antichi, Ran Ben Basat

机构 * Politecnico di Milano(米兰理工大学) Broadcom(博科公司)

AI总结 针对AI训练工作负载的网络与ML系统分离优化导致性能浪费问题,提出跨层联合优化的ML-for-ML方案,可使达到目标损失的速度提升最多42%。

Comments 8 pages, 3 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.00965 2026-08-04 cs.CR cs.AI 新提交

An AI Approach to Verified Production Cryptographic Libraries

验证型生产密码库的人工智能方法

Chuyue Sun, Su Fong, Zhiyi Kuang, Yizheng Jiao, Nina Narodytska, Haoze Wu, David L. Dill, Clark Barrett

机构 * Stanford University(斯坦福大学) University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校) VMware Research by Broadcom(博通旗下VMware研究院) Amherst College(阿默斯特学院)

AI总结 本文提出基于AI的CryptoProver系统,可从API合约合成内部规范与Verus验证的证明,成功验证了支撑Signal等系统的curve25519-dalek和RustCrypto的chacha20实现,遵循信任优先设计原则。

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.08923 2026-07-09 cs.LG cs.DC cs.NI 版本更新

DynamiQ: Accelerating Gradient Synchronization using Compressed Multi-hop All-reduce

DynamiQ:使用压缩多跳全规约加速梯度同步

Wenchen Han, Shay Vargaftik, Michael Mitzenmacher, Ran Ben Basat

机构 * University College London(伦敦大学) VMware Research by Broadcom(VMware由Broadcom进行的研究) Harvard University(哈佛大学) Broadcom

AI总结 研究针对大规模模型训练中网络成瓶颈的问题,提出DynamiQ量化框架,引入新技术并设计融合内核,扩展PyTorch DDP支持,在不同场景下比现有方法最多提升34.2%。

Comments 17 pages, 19 figures. Accepted to ACM SIGCOMM 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.15510 2026-06-24 cs.LG 版本更新

Not All Invariants Are Equal: Curating Training Data to Accelerate Program Verification with SLMs

并非所有不变式都同等重要:利用SLM通过精选训练数据加速程序验证

Ido Pinto, Yizhak Yisrael Elboher, Haoze Wu, Nina Narodytska, Guy Katz

机构 * Hebrew University of Jerusalem, Israel(特拉维夫大学) Amherst College, USA(阿默斯特学院) VMware Research by Broadcom, USA(Broadcom VMware 研究)

AI总结 提出Wonda数据筛选流水线,通过AST归一化和LLM语义重写从原始验证器输出中提取高质量训练不变式,微调小语言模型(SLM)后,在多个模型上使不变式正确性和加速率翻倍甚至三倍,小模型性能媲美大模型。

Comments A preprint of the ICML 2026 paper with the same title

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.04334 2026-05-13 cs.DB cs.AI cs.LO cs.PL

SpotIt+: Verification-based Text-to-SQL Evaluation with Database Constraints

SpotIt+: 基于验证的文本到SQL评估与数据库约束

Andrew Tremante, Yang He, Rocky Klopfenstein, Yuepeng Wang, Nina Narodytska, Haoze Wu

机构 * Amherst College(阿默斯特学院) Simon Fraser University(西蒙弗雷泽大学) VMware Research by Broadcom(Broadcom VMware 研究)

AI总结 SpotIt+通过有界等价验证评估文本到SQL系统,结合规则和LLM方法挖掘约束,提升数据库差异识别能力。

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.06014 2026-05-08 cs.LG cs.AI cs.DS cs.NI

Quantizing With Randomized Hadamard Transforms: Efficient Heuristic Now Proven

使用随机哈达玛变换进行量化:一种高效的启发式方法已被证明

Ran Ben-Basat, William Kuszmaul, Michael Mitzenmacher, Amit Portnoy, Shay Vargaftik

机构 * UCL and Broadcom(UCL和Broadcom) Carnegie Mellon University(卡内基梅隆大学) Harvard University(哈佛大学) Microsoft(微软) VMware Research by Broadcom(Broadcom的VMware研究)

AI总结 本文研究了随机哈达玛变换在量化中的应用,证明了使用两个变换可近似均匀随机旋转的效果,并展示了三个变换在向量量化中的优势。

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.18555 2026-05-08 cs.LG cs.AI cs.NI

A Note on TurboQuant and the Earlier DRIVE/EDEN Line of Work

关于TurboQuant及其早期DRIVE/EDEN工作的注记

Ran Ben-Basat, Yaniv Ben-Itzhak, Gal Mendelson, Michael Mitzenmacher, Amit Portnoy, Shay Vargaftik

机构 * UCL and Broadcom(UCL和博通) VMware Research by Broadcom(博通VMware研究) North Carolina State University(北卡罗来纳州立大学) Harvard University(哈佛大学) Microsoft(微软)

AI总结 本文分析了TurboQuant与DRIVE/EDEN在量化方法上的关系,指出TurboQuant_mse是EDEN的特例,而TurboQuant_prod在多个方面表现欠佳,EDEN在精度上优于TurboQuant。

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.26840 2026-03-05 cs.DB cs.AI cs.FL cs.LO

SpotIt: Evaluating Text-to-SQL Evaluation with Formal Verification

SpotIt: 通过形式验证评估文本到SQL

Rocky Klopfenstein, Yang He, Andrew Tremante, Yuepeng Wang, Nina Narodytska, Haoze Wu

机构 * Amherst College(阿姆赫斯特学院) Simon Fraser University(西蒙弗雷泽大学) VMware Research by Broadcom(VMware由Broadcom进行的研究)

AI总结 SpotIt通过形式验证评估文本到SQL系统,发现基于测试的方法无法区分生成查询与地面真相的差异。

Comments Accepted at ICLR'26

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.19132 2026-01-28 cs.NI cs.AI cs.AR cs.PF cs.SY eess.SY

In-Network Collective Operations: Game Changer or Challenge for AI Workloads?

网络内集体操作:人工智能工作负载中的变革者还是挑战?

Torsten Hoefler, Mikhail Khalilov, Josiah Clark, Surendra Anubolu, Mohan Kalkunte, Karen Schramm, Eric Spada, Duncan Roweth, Keith Underwood, Adrian Caulfield, Abdul Kabbani, Amirreza Rastegari

机构 * ETH Zürich(苏黎世联邦理工学院) Microsoft(微软) AMD(超威半导体) Broadcom Inc.(博通公司) Hewlett Packard Enterprise(惠普企业)

AI总结 本文探讨了网络内集体操作在人工智能工作负载中的潜力与挑战,分析了边缘和核心INC的性能优势及阻碍其应用的关键障碍,并预测其未来发展方向。

Journal ref IEEE Computer Jan. 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.19849 2026-01-23 cs.DC cs.AI cs.LG cs.NI

UCCL-EP: Portable Expert-Parallel Communication

UCCL-EP:便携式专家并行通信

Ziming Mao, Yihan Zhang, Chihan Cui, Zhen Huang, Kaichao You, Zhongjie Chen, Zhiying Xu, Zhenyu Gu, Scott Shenker, Costin Raiciu, Yang Zhou, Ion Stoica

机构 * UC Berkeley(加州大学伯克利分校) UC Davis(加州大学戴维斯分校) UW–Madison(威斯康星大学麦迪逊分校) AMD(超威半导体) Tsinghua University(清华大学) Amazon Web Services(亚马逊网络服务) ICSI(信息与计算机科学研究所) Broadcom & University Politehnica of Bucharest(博通与布加勒斯特理工大学)

AI总结 UCCL-EP通过高吞吐量GPU-CPU控制通道和RDMA立即数据模拟,实现了在异构GPU和NIC平台上的高性能专家并行通信,提升了多个模型的训练和推理效率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.01538 2026-01-08 cs.DB cs.CR cs.IR cs.LG

HONEYBEE: Efficient Role-based Access Control for Vector Databases via Dynamic Partitioning[Technical Report]

HONEYBEE: 通过动态分区实现向量数据库的高效基于角色的访问控制

Hongbin Zhong, Matthew Lentz, Nina Narodytska, Adriana Szekeres, Kexin Rong

机构 * Georgia Institute of Technology(佐治亚理工学院) Duke University(杜克大学) VMware Research by Broadcom(VMware由Broadcom进行的研究) Microsoft Research(微软研究院)

AI总结 HONEYBEE通过动态分区实现向量数据库的高效基于角色的访问控制,平衡内存开销与查询延迟,提升安全性与效率。

Comments Accepted by SIGMOD 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.19007 2025-12-23 cs.LG cs.AI

The 6th International Verification of Neural Networks Competition (VNN-COMP 2025): Summary and Results

第六届神经网络验证竞赛(VNN-COMP 2025):总结与结果

Konstantin Kaulen, Tobias Ladner, Stanley Bak, Christopher Brix, Hai Duong, Thomas Flinkow, Taylor T. Johnson, Lukas Koller, Edoardo Manino, ThanhVu H Nguyen, Haoze Wu

机构 * RWTH Aachen University(亚琛工业大学) Technical University of Munich(慕尼黑技术大学) Stony Brook University(石溪大学) George Mason University(乔治·梅奥大学) Maynooth University(梅诺特大学) Vanderbilt University(范德比尔特大学) University of Manchester(曼彻斯特大学) Amherst College(阿默斯特学院) VMware Research by Broadcom(Broadcom VMware 研究)

AI总结 VNN-COMP 2025通过标准化格式和公平评估,促进神经网络验证工具的比较与标准化。

Comments Report on the results of VNN-COMP 2025. arXiv admin note: substantial text overlap with arXiv:2412.19985, arXiv:2312.16760, arXiv:2212.10376

详情

展开后加载摘要…

URL PDF HTML 收藏
2501.07721 2025-01-15 cs.CL

LLMic: Romanian Foundation Language Model

Vlad-Andrei Bădoiu, Mihai-Valentin Dumitru, Alexandru M. Gherghescu, Alexandru Agache, Costin Raiciu

机构 * University Politehnica of Bucharest(布加勒斯特理工大学) Broadcom Inc.(博通公司)

详情

展开后加载摘要…

URL PDF HTML 收藏
2412.12170 2024-12-18 cs.LG cs.AI

PickLLM: Context-Aware RL-Assisted Large Language Model Routing

Dimitrios Sikeridis, Dennis Ramdass, Pranay Pareek

机构 * Broadcom(博通公司) VMware Cloud Foundation (VCF) Division(VMware 云基础架构(VCF)部门) AI and Advanced Services(AI 与高级服务部门)

Comments This work has been accepted at the first Workshop on Scalable and Efficient Artificial Intelligence Systems (SEAS) held in conjunction with the 39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025, in Philadelphia, Pennsylvania, USA

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.12819 2024-10-24 cs.DC cs.CL cs.LG cs.NI

I've Got 99 Problems But FLOPS Ain't One

Alexandru M. Gherghescu, Vlad-Andrei Bădoiu, Alexandru Agache, Mihai-Valentin Dumitru, Iuliu Vasilescu, Radu Mantu, Costin Raiciu

机构 * University Politehnica of Bucharest(布加勒斯特理工大学) Broadcom Inc.(博通公司)

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.14962 2024-08-26 cs.CL cs.AI cs.LG

Recent Advances in Generative AI and Large Language Models: Current Status, Challenges, and Perspectives

Desta Haileselassie Hagos, Rick Battle, Danda B. Rawat

机构 * Howard University(霍华德大学) VMware AI Labs by Broadcom(博通旗下VMware人工智能实验室) U.S. Army Research Laboratory (ARL)(美国陆军研究实验室)

Comments This version is accepted for publication in the Journal of IEEE Transactions on Artificial Intelligence (TAI)

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.13657 2024-07-19 cs.CL

FuLG: 150B Romanian Corpus for Language Model Pretraining

Vlad-Andrei Bădoiu, Mihai-Valentin Dumitru, Alexandru M. Gherghescu, Alexandru Agache, Costin Raiciu

机构 * University Politehnica of Bucharest(布加勒斯特理工大学) Broadcom Inc.(博通公司)

详情

展开后加载摘要…

URL PDF HTML 收藏
2405.04691 2024-05-09 cs.CR cs.LG

Carbon Filter: Real-time Alert Triage Using Large Scale Clustering and Fast Search

Jonathan Oliver, Raghav Batta, Adam Bates, Muhammad Adil Inam, Shelly Mehta, Shugao Xia

机构 * Broadcom Inc.(博通公司) University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)

详情

展开后加载摘要…

URL PDF HTML 收藏
2310.04870 2024-04-26 cs.FL cs.AI cs.LG cs.LO

Lemur: Integrating Large Language Models in Automated Program Verification

Haoze Wu, Clark Barrett, Nina Narodytska

机构 * Stanford University(斯坦福大学) VMware Research(威睿研究院) VMware by Broadcom(博通旗下威睿公司)

Comments Accepted at ICLR'24

详情

展开后加载摘要…

URL PDF HTML 收藏
2404.08865 2024-04-16 cs.CL cs.LG

LLM In-Context Recall is Prompt Dependent

Daniel Machlab, Rick Battle

机构 * VMware(威睿) Broadcom(博通)

详情

展开后加载摘要…

URL PDF HTML 收藏