Dongsheng Ding - Papers

Preprints

• Where to spend rollouts: Hit-utility optimal rollout allocation for group-based RLVR
  Tao Wang, Shuo Li, Yan Sun, Dongsheng Ding, Edgar Dobriban
  Note: arXiv:2605.07114v1; under review.

Refereed Papers

2026

• Convergence of natural policy gradient primal-dual methods for constrained convex MDPs
  Dongsheng Ding
  65th IEEE Conference on Decision and Control, 2026.
  Note: accepted and to appear.

• Unlearning in diffusion models: A unified framework with KL divergence and likelihood constraints
  Shervin Khalafi*, Alejandro Ribeiro, Dongsheng Ding*
  43rd International Conference on Machine Learning, 2026.
  Note: *corresponding authors; to appear; also arXiv:2605.30825.

2025

• Alignment of large language models with constrained learning
  Botong Zhang, Shuo Li, Ignacio Hounie, Osbert Bastani, Dongsheng Ding*, Alejandro Ribeiro
  Advances in Neural Information Processing Systems, 2025.
  Note: *corresponding author; poster; also arXiv:2505.19387 (oral presentation at COML Workshop).

• Composition and alignment of diffusion models using constrained learning
  Shervin Khalafi*, Ignacio Hounie, Dongsheng Ding*, Alejandro Ribeiro
  Advances in Neural Information Processing Systems, 2025.
  Note: *corresponding authors; poster; also arXiv:2508.19104.

• Convergence and sample complexity of natural policy gradient primal-dual methods for constrained MDPs
  Dongsheng Ding, Kaiqing Zhang, Jiali Duan, Tamer Basar, and Mihailo R. Jovanovic
  J. Mach. Learn. Res., 2025.
  Note: a journal extension of the paper; also arXiv:2206.02346.

• Deterministic policy gradient primal-dual methods for continuous-space constrained MDPs
  Sergio Rozada*, Dongsheng Ding*, Antonio G. Marques, Alejandro Ribeiro
  AAAI Conference on Artificial Intelligence, 2025.
  Note: *corresponding authors; poster; also arXiv:2408.10015.

2024

• Constrained diffusion models via dual training
  Shervin Khalafi, Dongsheng Ding*, and Alejandro Ribeiro
  Advances in Neural Information Processing Systems, 2024.
  Note: *corresponding author; poster.

• One-shot safety alignment for large language models via optimal dualization
  Xinmeng Huang, Shuo Li, Edgar Dobriban, Osbert Bastani, Hamed Hassani, Dongsheng Ding*
  Advances in Neural Information Processing Systems, 2024. (spotlight)
  Note: *corresponding author; poster; also video.

• Resilient constrained reinforcement learning
  Dongsheng Ding, Zhengyan Huan, and Alejandro Ribeiro
  27th International Conference on Artificial Intelligence and Statistics, 2024.

2023

• Last-iterate convergent policy gradient primal-dual methods for constrained MDPs
  Dongsheng Ding*, Chen-Yu Wei*, Kaiqing Zhang*, and Alejandro Ribeiro
  Advances in Neural Information Processing Systems, 2023.
  Note: *alphabetical order; poster, slides, video.

• Provably efficient generalized Lagrangian policy optimization for safe multi-agent reinforcement learning
  Dongsheng Ding, Xiaohan Wei, Zhuoran Yang, Zhaoran Wang, and Mihailo R. Jovanovic
  5th Annual Conference on Learning for Dynamics and Control, 2023.
  Note: poster.

2022

• Policy gradient primal-dual mirror descent for constrained MDPs with large state spaces
  Dongsheng Ding and Mihailo R. Jovanovic
  61st IEEE Conference on Decision and Control, 2022.
  Note: slides.

• Independent policy gradient for large-scale Markov potential games: sharper rates, function approximation, and game-agnostic convergence
  Dongsheng Ding*, Chen-Yu Wei*, Kaiqing Zhang*, and Mihailo R. Jovanovic
  39th International Conference on Machine Learning, 2022. (long talk)
  Note: *alphabetical order; poster, slides, video.

• Convergence and optimality of policy gradient primal-dual method for constrained Markov decision processes
  Dongsheng Ding, Kaiqing Zhang, Tamer Basar, and Mihailo R. Jovanovic
  2022 American Control Conference, 2022.
  Note: slides.

2021

• Provably efficient safe exploration via primal-dual policy optimization
  Dongsheng Ding, Xiaohan Wei, Zhuoran Yang, Zhaoran Wang, and Mihailo R. Jovanovic
  24th International Conference on Artificial Intelligence and Statistics, 2021. (oral presentation)
  Note: poster, slides, video.

• Byzantine-resilient distributed learning under constraints
  Dongsheng Ding, Xiaohan Wei, Hao Yu, and Mihailo R. Jovanovic
  2021 American Control Conference, 2021.
  Note: slides.

• Discounted online Newton method for time-varying time series prediction
  Dongsheng Ding, Jianjun Yuan, and Mihailo R. Jovanovic
  2021 American Control Conference, 2021.
  Note: slides.

2020

• Natural policy gradient primal-dual method for constrained Markov decision processes
  Dongsheng Ding, Kaiqing Zhang, Tamer Basar, and Mihailo R. Jovanovic
  Advances in Neural Information Processing Systems, 2020.
  Note: poster, slides, video.

• Global exponential stability of primal-dual gradient flow dynamics based on the proximal augmented Lagrangian: A Lyapunov-based approach
  Dongsheng Ding and Mihailo R. Jovanovic
  59th IEEE Conference on Decision and Control, 2020.
  Note: slides.

2019

• Fast multi-agent temporal-difference learning via homotopy stochastic primal-dual method
  Dongsheng Ding, Xiaohan Wei, Zhuoran Yang, Zhaoran Wang, and Mihailo R. Jovanovic
  Optimization Foundations for Reinforcement Learning Workshop, 2019.
  Note: poster; also arXiv:1908.02805.

• Distributed robust statistical learning: Byzantine mirror descent
  Dongsheng Ding, Xiaohan Wei, and Mihailo R. Jovanovic
  58th IEEE Conference on Decision and Control, 2019
  Note: poster, slides.

• Global exponential stability of primal-dual gradient flow dynamics based on the proximal augmented Lagrangian
  Dongsheng Ding and Mihailo R. Jovanovic
  2019 American Control Conference, 2019.
  Note: slides.

2018

• An exponentially convergent primal-dual algorithm for nonsmooth composite minimization
  Dongsheng Ding, Bin Hu, Neil K. Dhingra, and Mihailo R. Jovanovic
  57th IEEE Conference on Decision and Control, 2018.
  Note: slides.

• A primal-dual Laplacian gradient flow dynamics for distributed resource allocation problems
  Dongsheng Ding and Mihailo R. Jovanovic
  2018 American Control Conference, 2018.
  Note: slides.

2016 and earlier

• Adaptive Mittag-Leffler stabilization of a class of fractional order uncertain nonlinear systems
  Qiao Wang, Jianliang Zhang, Dongsheng Ding, and Donglian Qi
  Asian J. Control, 2016.

• Nonlinear Mittag-Leffler stabilization of commensurate fractional-order nonlinear systems
  Dongsheng Ding, Donglian Qi, and Qiao Wang
  IET Control Theory Appl., 2015

• Asymptotic pseudo-state stabilization of uncertain fractional-order nonlinear systems with additive disturbance
  Dongsheng Ding, Donglian Qi, and Qiao Wang
  Nonlinear Dyn., 2015.

• Mittag-Leffler synchronization of uncertain fractional order chaotic systems
  Qiao Wang, Dongsheng Ding, and Donglian Qi
  Chinese Physics B, 2015.

• Adaptive Mittag-Leffler stabilization of commensurate fractional-order nonlinear systems
  Dongsheng Ding, Donglian Qi, Yao Meng, and Li Xu
  53rd IEEE Conference on Decision and Control, 2014.

• Strategy analysis of an evolutionary spectrum sensing game
  Dongsheng Ding, Guoyue Zhang, Donglian Qi, and Huhu Zhang
  Intelligent Computing and Applications (LSMS & ICSEE), 2014.

• Alternative LMI characterizations for fractional-order linear systems
  Dongsheng Ding, Donglian Qi, and Qiao Wang
  33rd Chinese Control Conference, 2014.

• Fractional-order integral state space modeling and quasi state analysis via block operational matrix scheme
  Dongsheng Ding, Donglian Qi, and Qiao Wang
  26th Chinese Control and Decision Conference, 2014.

• Convergence analysis and performance of an extended central force optimization algorithm
  Dongsheng Ding, Donglian Qi, Xiaoping Luo, Jinfei Chen, Xuejie Wang, and Pengying Du
  Appl. Math. Comput., 2012

• A convergence proof and parameter analysis of central force optimization algorithm
  Dongsheng Ding, Xiaoping Luo, Jinfen Chen, Xuejie Wang, Pengying Du, and Yunfei Guo
  J. Convergence Inf. Technol., 2011.