arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

NeurIPS

Conference on Neural Information Processing Systems · 会议 · Machine Learning

2026-05-14 至 2026-05-14 共收录 15
2605.13716 2026-05-14 cs.SE cs.MA

SkillOps: Managing LLM Agent Skill Libraries as Self-Maintaining Software Ecosystems

SkillOps:将LLM代理技能库视为自维护的软件生态系统进行管理

Hongji Pu, Xinyuan Song, Liang Zhao

AI总结 SkillOps通过自维护软件生态系统方法管理LLM代理技能库,解决技能库中的技术债务问题,提升技能检索、组合和执行效率,实验表明其在ALFWorld上任务成功率高达79.5%。

Comments 23 pages, 9 figures. Submitted to NeurIPS 2026. Code is available at https://github.com/Hik289/SkillOps.git

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.13638 2026-05-14 quant-ph cs.LG

CO-MAP: A Reinforcement Learning Approach to the Qubit Allocation Problem

CO-MAP:一种用于量子比特分配问题的强化学习方法

Ankit Kulshrestha, Xiaoyuan Liu

机构 * Fujitsu Research of America(富士通美国研究院)

AI总结 本文提出CO-MAP方法,通过强化学习策略优化量子比特映射,显著减少SWAP门开销,实测数据表明在MQTBench和Queko电路中SWAP开销降低65-85%。

Comments Under review at NeurIPS'26

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.13386 2026-05-14 cs.LG stat.ML

Support-Conditioned Flow Matching Is Kernel Smoothing

支持条件流匹配是核平滑

Daniel Matsui Smola

机构 * Department of Computer Science(计算机科学系) University of Washington(华盛顿大学)

AI总结 本文通过核平滑理论揭示流匹配中支持集的核密度估计机制,证明流时间影响核带宽,并验证跨注意力在高维数据中的局限性。

Comments Submitted to NeurIPS 2026. 18 pages, 10 figures, 1 table. Code at https://github.com/BaroqueObama/kernel-flow-matching-code

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.13384 2026-05-14 cs.LG

Teaching and Learning under Deductive Errors

在演绎错误下教学与学习

Jan Arne Telle, Brigt Håvardstun, Jose Hernandez-Orallo

机构 * Department of Informatics University of Bergen(卑尔根大学信息学院) University of Bergen(卑尔根大学) VRAIN - Universitat Politecnica de Valencia(瓦伦西亚理工大学VRAIN实验室) Universitat Politecnica de Valencia(瓦伦西亚理工大学) Leverhulme Centre for the Future of Intelligence - University of Cambridge(剑桥大学未来智能中心)

AI总结 本文研究了在存在演绎错误的教学与学习框架,分析了机器教学中如何通过PAC设定生成近似正确的假设,并探讨了计算复杂性及实验验证。

Comments 15 pages, preprint neurips

详情

展开后加载摘要…

URL PDF HTML 收藏
2302.14234 2026-05-14 cs.GT econ.TH

Bicriteria Multidimensional Mechanism Design with Side Information

双目标多维机制设计与侧信息

Maria-Florina Balcan, Siddharth Prasad, Tuomas Sandholm

AI总结 本文提出一种多维机制设计方法,利用侧信息提升效率与收益。通过整合改进的VCG机制与最弱类型代理,证明在高质量侧信息下,机制性能可与无先验总社会效益竞争。

Comments Mathematics of Operations Research Special Issue on Market Design; supersedes the NeurIPS 2023 paper of the same title

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.13188 2026-05-14 stat.ML cs.CL cs.LG stat.ME

LLMs as Implicit Imputers: Uncertainty Should Scale with Missing Information

大型语言模型作为隐式填补器:不确定性应与缺失信息成比例

Stef van Buuren

机构 * TNO - Netherlands Organization for Applied Scientific Research(荷兰应用科学研究院) Dept. of Methodology and Statistics, University of Utrecht(乌得勒支大学方法学与统计学系)

AI总结 本文探讨了大型语言模型在不完整上下文下的不确定性衡量,通过实验发现熵比置信度更能反映缺失信息的影响,且在不同证据水平上解释了更多准确性变化。

Comments 9 pages, 3 figures, 2 tables, NeurIPS 2026 position paper

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.12999 2026-05-14 q-bio.NC cs.LG

Implicit Behavioral Decoding from Next-Step Spike Forecasts at Population Scale

从群体尺度上的下一步尖峰预测隐式行为解码

John R. Minnick, Jesus Gonzalez-Ferrer, Kamran Hussain, Jinghui Geng, Ash Robbins, Mohammed A. Mostajo-Radji, David Haussler, Jason Eshraghian, Mircea Teodorescu

机构 * Department of Electrical and Computer Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校电气与计算机工程系) UC Santa Cruz Genomics Institute, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校基因组研究所) Department of Biomolecular Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校生物分子工程系) Department of Applied Mathematics, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校应用数学系) Department of Computer Science and Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校计算机科学与工程系)

AI总结 本文提出利用单个Mamba预测器在一次前向传递中实现神经群体活动预测和动物行为状态解码,通过轻量级线性头在匹配时间上下文中优于原始尖峰计数解码,验证了在Steinmetz视觉辨别基准上的有效性。

Comments 21 pages, 6 figures, 5 tables; submitted to NeurIPS 2026 Neuroscience & Cognitive Science Track

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.12992 2026-05-14 q-bio.NC cs.LG

SpikeProphecy: A Large-Scale Benchmark for Autoregressive Neural Population Forecasting

SpikeProphecy:大规模用于自回归神经群体预测的基准测试

John R. Minnick, Jinghui Geng, Kamran Hussain, Jesus Gonzalez-Ferrer, Ash Robbins, Mohammed A. Mostajo-Radji, David Haussler, Jason K. Eshraghian, Mircea Teodorescu

机构 * Department of Electrical and Computer Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校电气与计算机工程系) UC Santa Cruz Genomics Institute, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校基因组研究所) Department of Computer Science and Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校计算机科学与工程系) Department of Applied Mathematics, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校应用数学系) Department of Biomolecular Engineering, University of California, Santa Cruz, CA, USA(加州大学圣克ruz分校生物分子工程系)

AI总结 本文提出SpikeProphecy基准测试,用于评估自回归神经群体预测,通过分解指标揭示数据结构,验证了不同架构在预测脑区可预测性上的表现。

Comments 26 pages, 4 figures, 12 tables; submitted to NeurIPS 2026 Datasets and Benchmarks Track; processed dataset at https://huggingface.co/datasets/mysteriousauthor/spikeprophecy-steinmetz (CC-BY-4.0); code at https://github.com/JohnMinnick/SpikeProphecy-A-Large-Scale-Benchmark-for-Autoregressive-Neural-Population-Forecasting

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.12736 2026-05-14 cs.LG

ConRetroBert: EMA Stabilized Dual Encoders for Template-Based Single-Step Retrosynthesis

ConRetroBert:基于指数移动平均的双编码器用于基于模板的单步逆合成

Mohammad Jahid Ibna Basher, Ali Khodabandeh Yalabadi, Ivan Garibay, Ozlem Ozmen Garibay

机构 * Department of Industrial Engineering(工业工程系)

AI总结 ConRetroBert通过将模板逆合成重构为密集产品模板检索和候选集列表排序,提升了模板基于方法的性能,其双编码器框架在USPTO-50k基准上实现了更高的反应准确率。

Comments Submitted to NeurIPS 2026 Main Conference

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.12693 2026-05-14 cs.LG

IGT-OMD: Implicit Gradient Transport for Decision-Focused Learning under Delayed Feedback

IGT-OMD:隐式梯度传输用于延迟反馈下的决策聚焦学习

Benjamin Amoh, Geoffrey G. Parker, Wesley Marrero

机构 * Thayer School of Engineering, Dartmouth College(达特茅斯学院泰勒工程学院)

AI总结 本文提出IGT-OMD算法,通过隐式梯度传输减少延迟反馈下的传输误差,实现子线性 regrets 绑定,并在多个任务中验证了其有效性。

Comments 9 pages, 4 figures, NeurIPS 2026 conference

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.04759 2026-05-14 cs.CL cs.AI cs.ET cs.LG

Gyan: An Explainable Neuro-Symbolic Language Model

Gyan:一种可解释的神经符号语言模型

Venkat Srinivasan, Vishaal Jatav, Anushka Chandrababu, Geetika Sharma

机构 * Innospark Ventures & Gyan AI(Innospark Ventures及Gyan AI) Gyan AI Inc.(Gyan AI公司)

AI总结 Gyan基于非Transformer架构构建,克服了传统大语言模型的局限性,在多个数据集上取得SOTA表现,展示了可信任且可靠的使命关键任务模型潜力。

Comments also submitted to NeurIPS 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.08944 2026-05-14 cs.LG cs.MA

Multi-Agent Decision-Focused Learning via Value-Aware Sequential Communication

多智能体决策导向学习 via 值感知序列通信

Benjamin Amoh, Geoffrey Parker, Wesley Marrero

机构 * Thayer School of Engineering, Dartmouth College(达特茅斯大学泰勒工程学院)

AI总结 本文提出SeqComm-DFL方法,通过值感知序列通信与决策导向学习结合,提升多智能体任务性能。方法引入序列Stackelberg条件生成消息,利用信息论界限证明收敛性,并在协作医疗和StarCraft多智能体挑战中取得显著奖励和胜率提升。

Comments 9 pages, 2 figues, 1 table, neurips 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.15743 2026-05-14 cs.LG astro-ph.EP astro-ph.IM

Connecting the Dots: A Machine Learning Ready Dataset for Ionospheric Forecasting Models

连接点:为电离层预报模型准备的机器学习数据集

Linnea M. Wolniewicz, Halil S. Kelebek, Simone Mestici, Michael D. Vergalla, Giacomo Acciarini, Bala Poduval, Olga Verkhoglyadova, Madhulika Guhathakurta, Thomas E. Berger, Atılım Güneş Baydin, Frank Soboczenski

机构 * Department of Information and Computer Science(信息与计算机科学系) University of Hawai‘i at Mānoa(夏威夷大学毛纳罗亚分校) Department of Engineering Science(工程科学系) University of Oxford(牛津大学) Università degli Studi di Roma Sapienza(罗马大学) Free Flight Research Lab(自由飞行研究实验室) University of New Hampshire(新罕布什尔大学) European Space Agency (ESA)(欧洲航天局) NASA Jet Propulsion Laboratory(美国宇航局喷气推进实验室) NASA Headquarters(美国宇航局总部) Space Weather Technology, Research, and Education Center(空间天气技术、研究与教育中心) University of Colorado Boulder(科罗拉多大学博尔德分校) Department of Computer Science(计算机科学系) University of York & King’s College London(约克大学及伦敦国王学院)

AI总结 本文提出一个整合多种电离层和日球层数据的机器学习数据集,用于改进电离层预报模型,支持科学探索和实际应用。

Comments 8 pages, 2 figures, 2 tables. Accepted as a poster presentation in the Machine Learning for the Physical Sciences workshop at NeurIPS 2025. Dataset can be found on Zenodo (https://zenodo.org/records/18343833) or GitHub (https://github.com/FrontierDevelopmentLab/2025-HL-Ionosphere-dataset)

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.21060 2026-05-14 cs.LG cs.AI

On the Sample Complexity of Differentially Private Policy Optimization

关于差分隐私策略优化的样本复杂性

Yi He, Xingyu Zhou

机构 * Wayne State University(韦恩州立大学)

AI总结 本文研究了差分隐私策略优化的样本复杂性,分析了多种策略优化算法在隐私约束下的样本复杂性,揭示隐私成本在样本复杂性中的表现。

Comments Accepted at NeurIPS 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.01502 2026-05-14 q-bio.NC cs.CV cs.LG

Behavioral Geometric Supervision Aligns Video Foundation Models with Human Social Perception

行为几何监督使视频基础模型与人类社会感知对齐

Kathy Garcia, Leyla Isik

机构 * Department of Cognitive Science(认知科学系) Department of Biomedical Engineering(生物医学工程系) Johns Hopkins University(约翰霍普金斯大学)

AI总结 本文提出行为几何监督(BGS),通过引入人类社会判断数据,提升视频模型对社会关系的感知能力,实验表明该方法能显著提升模型性能并揭示可解释的社会情感属性。

Comments v2: Major revision. Retitled; expanded from TimeSformer alone to four backbones (V-JEPA 2/2.1, TimeSformer, VideoMAE, CLIP), with V-JEPA 2.1 nearly tripling pretrained performance. Adds zero-shot PHASE transfer, attention-rollout analysis, and a language-distillation control. Data (OOO sim. judgments) & core hybrid triplet+RSA LoRA method unchanged from v1. Prepared for NeurIPS 2026 submission

详情

展开后加载摘要…

URL PDF HTML 收藏