世界模型 - arXivDaily 专题

2309.02873 2024-01-31 cs.LG 53%

Learning Hybrid Dynamics Models With Simulator-Informed Latent States

Katharina Ensinger, Sebastian Ziesche, Sebastian Trimpe

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Comments Accepted at The 38th Annual AAAI Conference on Artificial Intelligence, 2024

2312.01779 2023-12-05 cs.MA physics.soc-ph 53%

Viral transmission in pedestrian crowds: Coupling an open-source code assessing the risks of airborne contagion with diverse pedestrian dynamics models

Alexandre Nicolas, Simon Mendez

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.MA

2306.09182 2023-06-16 cs.RO 53%

Rolling control and dynamics model of two section articulated-wing ornithopter

G. Su, Y. Cai, J. Zhao

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.RO

2211.12921 2022-12-07 cs.RO cs.SY eess.SY 53%

Hybrid Learning of Time-Series Inverse Dynamics Models for Locally Isotropic Robot Motion

Tolga-Can Çallar, Sven Böttger

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.RO

Comments Accepted for publication in IEEE Robotics and Automation Letters ( see https://ieeexplore.ieee.org/document/9954138 ). 8 pages, 8 figures

2210.04958 2022-10-14 cs.LG stat.ME 53%

Mining Causality from Continuous-time Dynamics Models: An Application to Tsunami Forecasting

Fan Wu, Sanghyun Hong, Donsub Rim, Noseong Park, Kookjin Lee

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

2205.04796 2022-05-11 cs.RO 53%

Efficient Learning of Inverse Dynamics Models for Adaptive Computed Torque Control

David Jorge, Gabriella Pizzuto, Michael Mistry

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.RO

Comments Submitted to IEEE/RSJ International Conference on Intelligent Robots & Systems (IROS) 2022

2106.11609 2021-10-18 cs.LG math.DS stat.ML 53%

Distributional Gradient Matching for Learning Uncertain Neural Dynamics Models

Lenart Treven, Philippe Wenk, Florian Dörfler, Andreas Krause

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Comments Published at NeurIPS 2021

Journal ref Advances in Neural Information Processing Systems, 2021

2006.08935 2021-06-08 cs.LG math.DS stat.ML 53%

Learning Dynamics Models with Stable Invariant Sets

Naoya Takeishi, Yoshinobu Kawahara

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Journal ref Proceedings of the AAAI Conference on Artificial Intelligence, 35(11), 9782-9790, 2021

2104.07252 2021-04-22 cs.AI cs.CL 53%

Emotion Dynamics Modeling via BERT

Haiqin Yang, Jianping Shen

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.AI

Comments 9 pages, 9 figures, 5 tables, in IJCNN 2021

2011.11751 2021-03-31 cs.RO 53%

Multimodal dynamics modeling for off-road autonomous vehicles

Jean-François Tremblay, Travis Manderson, Aurélio Noca, Gregory Dudek, David Meger

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.RO

2004.14548 2020-05-01 physics.soc-ph cs.MA nlin.AO 53%

Extremism definitions in opinion dynamics models

André C. R. Martins

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.MA

Comments 17 pages, 2 figures, 1 table, to appear in the ICCS2020 Proceedings (Tenth International Conference on Complex Systems)

1910.04297 2020-01-29 cs.RO cs.LG 53%

Online Simultaneous Semi-Parametric Dynamics Model Learning

Joshua Smith, Michael Mistry

专题命中模型式强化学习：dynamics model(title)；分类 cs.LG、cs.RO

Comments \c{opyright} 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

2001.06116 2020-01-20 cs.LG math.DS stat.ML 53%

Learning Stable Deep Dynamics Models

Gaurav Manek, J. Zico Kolter

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Comments NeurIPS 2019

1911.03318 2019-11-11 stat.ML cs.LG cs.SY eess.SY 53%

Deep Transfer Learning for Thermal Dynamics Modeling in Smart Buildings

Zhanhong Jiang, Young M. Lee

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Comments 5 pages, 2 figures; Accepted at 2019 IEEE International Conference on Big Data (IEEE BigData 2019)

1809.05074 2018-09-14 cs.LG stat.ML 53%

Derivative-free online learning of inverse dynamics models

Diego Romeres, Mattia Zorzi, Raffaello Camoriano, Silvio Traversaro, Alessandro Chiuso

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

Comments 14 pages, 11 figures

1603.05412 2016-10-11 math.OC cs.LG stat.ML 53%

Online semi-parametric learning for inverse dynamics modeling

Diego Romeres, Mattia Zorzi, Raffaello Camoriano, Alessandro Chiuso

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.LG

1010.3361 2015-03-17 q-bio.PE cond-mat.dis-nn cond-mat.stat-mech math-ph math.MP 53%

Full analytical solution and complete phase diagram analysis of the Verhulst-like two-species population dynamics model

Brenno Caetano Troca Cabella, Alexandre Souto Martinez, Fabiano Ribeiro

专题命中模型式强化学习：dynamics model(title,abstract)；simulation model(abstract)

Comments 12 pages and 4 figures

1309.3660 2013-09-17 cs.SI cs.MA nlin.AO physics.soc-ph 53%

(Failure of the) Wisdom of the crowds in an endogenous opinion dynamics model with multiply biased agents

Steffen Eger

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.MA

Comments 56 pages, 17 figures

1004.5215 2010-07-05 cs.AI q-bio.CB 53%

System Dynamics Modelling of the Processes Involving the Maintenance of the Naive T Cell Repertoire

Grazziela P. Figueredo, Uwe Aickelin, Amanda Whitbrook

专题命中模型式强化学习：dynamics model(title,abstract)；分类 cs.AI

Comments 6 pages, 2 figures, 1 table, 9th Annual Workshop on Computational Intelligence (UKCI 2009), Nottingham, UK

Journal ref Proceedings of the 9th Annual Workshop on Computational Intelligence (UKCI 2009), Nottingham, UK, p13-18,

2605.12763 2026-08-19 cs.LG math.DS math.OC q-bio.NC 版本更新 50%

Center-Manifold Reduction of Learning at Bifurcations: Interference and Rich Learning in Recurrent Neural Networks

状态空间NTK在接近分岔点时的坍缩

James Hazelden, Eric Shea-Brown

机构 * University of Washington（华盛顿大学）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结研究通过经验状态空间神经 tangent 核（sNTK）分析梯度下降在分岔点附近的动态，发现分岔点主导并简化了学习过程，通过分解sNTK揭示了高维递归系统的学习几何结构。

2602.06323 2026-08-19 cs.LG 版本更新 50%

How (Not) to Hybridize Neural and Mechanistic Models for Epidemiological Forecasting

如何（不）将神经模型与机制模型混合用于流行病预测

Yiqi Su, Ray Lee, Jiaming Cui, Naren Ramakrishnan

机构 * Virginia Tech（弗吉尼亚理工大学）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结针对流行病预测中神经-机制混合模型在部分可观测和非平稳动态下失效的问题，提出通过显式分解感染序列的多尺度结构作为控制信号，驱动受控神经ODE与流行病模型耦合，实现最优预测和时变参数推断。

2608.12334 2026-08-14 cs.CL cs.AI 新提交 50%

Steering the Language Axis: From Linear Decodability to Causal Control

操控语言轴：从线性可解码到因果控制

Arnav Srivastav

机构 * University of California, Santa Cruz（加利福尼亚大学圣克鲁兹分校）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.AI

AI总结本研究探究大语言模型的语言身份是否可通过紧凑激活方向因果控制，在Qwen、Llama等模型的126万次生成上证实，分离的PCA导出语言轴可可靠操控语言切换，且语言选择具层特异性与语言对依赖性。

Comments 22 pages, 14 figures, Under review

2601.19612 2026-08-14 cs.LG cs.AI cs.RO 版本更新 50%

Safe Exploration via Policy Priors

通过策略先验进行安全探索

Manuel Wendl, Yarden As, Manish Prajapat, Anton Pollak, Stelian Coros, Andreas Krause

机构 * ETH Zurich（苏黎世联邦理工学院）

专题命中模型式强化学习：分类 cs.AI、cs.LG、cs.RO；dynamics model(abstract)

AI总结提出SOOPER方法，利用次优但保守的策略先验，结合概率动力学模型进行乐观探索和悲观回退，在保证安全的同时收敛到最优策略。

2606.02363 2026-08-11 cs.LG stat.ML 版本更新 50%

Minimax-Optimal Policy Regret in Partially Observable Markov Games

部分可观测马尔可夫博弈中的极小化最优策略遗憾

Raman Arora

机构 * Raman Arora

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结针对部分可观测马尔可夫博弈，提出基于epoch的乐观最大似然算法，实现了与聚合Eluder维数相关的$ ilde{O}(\sqrt{T})$策略遗憾，并证明了匹配的下界。

2604.00669 2026-08-10 cs.LG math.DS 版本更新 50%

Embedded Variational Neural Stochastic Differential Equations for Learning Heterogeneous Dynamics

嵌入变分神经随机微分方程用于学习异质动态

Sandeep Kumar Samota, Reema Gupta, Snehashish Chakraverty

机构 * National Institute of Technology, Rourkela, India（印度国立理工学院鲁尔克拉分校）； Poverty Alleviation Research Centre, National Institute of Technology, Rourkela, India（印度国立理工学院鲁尔克拉分校扶贫研究中心）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结本文提出V-NSDE模型，结合神经SDE的动态表达与VAE的生成能力，用于学习异质动态，通过编码器和解码器处理时间序列数据，提升复杂模式识别能力。

2601.15363 2026-07-24 stat.ML cs.LG 版本更新 50%

Non-Stationary Functional Bilevel Optimization

非平稳函数双层优化

Jason Bohne, Ieva Petrulionyte, Michael Arbel, Julien Mairal, Paweł Polak

机构 * Applied Mathematics and Statistics, Stony Brook University, Stony Brook, NY, USA（应用数学与统计学系，史泰文布鲁克大学）； Inria, CNRS, Grenoble INP, LJK, Université Grenoble Alpes（法国国家信息与自动化技术研究院、国家科学研究中心、格勒诺布尔INP、LJK、格勒诺布尔阿尔卑斯大学）

专题命中模型式强化学习：model-based reinforcement learning(abstract)；分类 cs.LG

AI总结 SmoothFBO是首个在非平稳函数双层优化中具有理论保证和实际可扩展性的算法，通过时间平滑的随机超梯度估计器提升稳定性和性能。

2510.01475 2026-07-20 eess.SY cs.LG cs.SY 版本更新 50%

Comparative Field Deployment of Reinforcement Learning and Model Predictive Control for Residential HVAC

强化学习与模型预测控制在住宅暖通空调中的现场对比部署

Ozan Baris Mulayim, Elias N. Pergantis, Levi D. Reyes Premer, Bingqing Chen, Guannan Qu, Kevin J. Kircher, Mario Bergés

机构 * College of Engineering, Carnegie Mellon University, 5000 Forbes Ave, Pittsburgh, PA 15213, USA ； Wilton E. Scott Institute for Energy Innovation, Carnegie Mellon University, 5000 Forbes Ave, Pittsburgh, PA 15213, USA ； Center for High Performance Buildings, Purdue University, 177 S Russell St, West Lafayette, IN 47907, USA ； Trane Technologies, Residential R\&D Group, 6200 Troup Hwy, Tyler, TX 75707, USA ； Bosch Center for Artificial Intelligence

专题命中模型式强化学习：model-based RL(abstract)；分类 cs.LG

AI总结研究对比强化学习与模型预测控制应用于住宅暖通空调的情况，通过在寒冷气候住宅中各部署一个月，对比两者节能效果、居住者舒适度及数据需求等，发现RL节能略优、部署工作量少但面临初始化等困难。

Comments 26 pages, 9 figures, 5 tables. Under review for Applied Energy

2607.09801 2026-07-14 cs.LG 新提交 50%

Discovering Latent Response Laws in Forced Physical Systems

在强迫物理系统中发现潜在响应规律

Yi Zhu, Su Chen, Xiaojun Li, Xiuli Du

机构 * Beijing University of Technology（北京工业大学）； State Key Laboratory of Bridge Safety and Resilience（桥梁安全与韧性国家重点实验室）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结研究强迫物理系统中潜在响应规律，提出FLARE方法，通过学习紧凑响应坐标等，能恢复强迫动力学并预测高维响应，还可将方程发现扩展到复杂观测系统，为可解释建模和预测提供途径。

Comments 23 pages, 5 figures and 2 tables. Supplementary Information provided as an ancillary file

2607.10965 2026-07-14 physics.comp-ph math-ph math.MP physics.data-an physics.flu-dyn 新提交 50%

Structure-preserving variational neural fields: Uncertainty-quantified reduced-order modeling of nonlinear conservation laws

结构保持变分神经场：非线性守恒律的不确定性量化降阶建模

Aviral Prakash, Marc L. Klasky

专题命中模型式强化学习：latent dynamics(abstract)；dynamics model(abstract)

AI总结研究针对非线性守恒律控制的物理系统模拟，开发变分潜在神经场框架，集成高斯过程代理模型，有三种变体，能估计预测置信度，通过时空无散度表示嵌入守恒律流形实现守恒结构保持，经实验验证其有效性和鲁棒性。

2607.01022 2026-07-02 cs.LG 新提交 50%

Seahorse: A Unified Benchmarking Framework for Spatiotemporal Event Modeling

Seahorse: 时空事件建模的统一基准框架

Yahya Aalaila, Gerrit Großmann, Sebastian Vollmer

机构 * German Research Center for Artificial Intelligence (DFKI), Data Science and its Applications Research Group, Kaiserslautern, Germany（德国人工智能研究中心（DFKI），数据科学及其应用研究组，凯泽斯劳滕，德国）； Department of Computer Science, Rhineland-Palatinate Technical University of Kaiserslautern-Landau (RPTU), Kaiserslautern, Germany（计算机科学系，莱茵兰-普法尔茨凯泽斯劳滕-兰道工业大学（RPTU），凯泽斯劳滕，德国）

专题命中模型式强化学习：latent dynamics(abstract)；分类 cs.LG

AI总结提出Seahorse统一框架，通过编码-演化-解码接口标准化神经时空点过程，实现公平比较与诊断分析，并引入合成压力测试套件揭示各模型族的归纳偏差。

Comments 24 pages, 9 figures. Code: https://github.com/YahyaAalaila/seahorse

1. 模型式强化学习 1124 篇

Learning Hybrid Dynamics Models With Simulator-Informed Latent States

Viral transmission in pedestrian crowds: Coupling an open-source code assessing the risks of airborne contagion with diverse pedestrian dynamics models

Rolling control and dynamics model of two section articulated-wing ornithopter

Hybrid Learning of Time-Series Inverse Dynamics Models for Locally Isotropic Robot Motion

Mining Causality from Continuous-time Dynamics Models: An Application to Tsunami Forecasting

Efficient Learning of Inverse Dynamics Models for Adaptive Computed Torque Control

Distributional Gradient Matching for Learning Uncertain Neural Dynamics Models

Learning Dynamics Models with Stable Invariant Sets

Emotion Dynamics Modeling via BERT

Multimodal dynamics modeling for off-road autonomous vehicles

Extremism definitions in opinion dynamics models

Online Simultaneous Semi-Parametric Dynamics Model Learning

Learning Stable Deep Dynamics Models

Deep Transfer Learning for Thermal Dynamics Modeling in Smart Buildings

Derivative-free online learning of inverse dynamics models

Online semi-parametric learning for inverse dynamics modeling

Full analytical solution and complete phase diagram analysis of the Verhulst-like two-species population dynamics model

(Failure of the) Wisdom of the crowds in an endogenous opinion dynamics model with multiply biased agents

System Dynamics Modelling of the Processes Involving the Maintenance of the Naive T Cell Repertoire

Center-Manifold Reduction of Learning at Bifurcations: Interference and Rich Learning in Recurrent Neural Networks

How (Not) to Hybridize Neural and Mechanistic Models for Epidemiological Forecasting

Steering the Language Axis: From Linear Decodability to Causal Control

Safe Exploration via Policy Priors

Minimax-Optimal Policy Regret in Partially Observable Markov Games

Embedded Variational Neural Stochastic Differential Equations for Learning Heterogeneous Dynamics

Non-Stationary Functional Bilevel Optimization

Comparative Field Deployment of Reinforcement Learning and Model Predictive Control for Residential HVAC

Discovering Latent Response Laws in Forced Physical Systems

Structure-preserving variational neural fields: Uncertainty-quantified reduced-order modeling of nonlinear conservation laws

Seahorse: A Unified Benchmarking Framework for Spatiotemporal Event Modeling