Saved in:
Bibliographic Details
Main Authors: Zhang, Lunjun, Han, Shuo, Lyu, Hanrui, Stadie, Bradly C
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.03508
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910247037173760
author Zhang, Lunjun
Han, Shuo
Lyu, Hanrui
Stadie, Bradly C
author_facet Zhang, Lunjun
Han, Shuo
Lyu, Hanrui
Stadie, Bradly C
contents We introduce D2AC, a new model-free reinforcement learning (RL) algorithm designed to train expressive diffusion policies online effectively. At its core is a policy improvement objective that avoids the high variance of typical policy gradients and the complexity of backpropagation through time. This stable learning process is critically enabled by our second contribution: a robust distributional critic, which we design through a fusion of distributional RL and clipped double Q-learning. The resulting algorithm is highly effective, achieving state-of-the-art performance on a benchmark of eighteen hard RL tasks, including Humanoid, Dog, and Shadow Hand domains, spanning both dense-reward and goal-conditioned RL scenarios. Beyond standard benchmarks, we also evaluate a biologically motivated predator-prey task to examine the behavioral robustness and generalization capacity of our approach. Code: https://github.com/d2ac-actor-critic/d2ac-public
format Preprint
id arxiv_https___arxiv_org_abs_2510_03508
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle D2 Actor Critic: Diffusion Actor Meets Distributional Critic
Zhang, Lunjun
Han, Shuo
Lyu, Hanrui
Stadie, Bradly C
Machine Learning
We introduce D2AC, a new model-free reinforcement learning (RL) algorithm designed to train expressive diffusion policies online effectively. At its core is a policy improvement objective that avoids the high variance of typical policy gradients and the complexity of backpropagation through time. This stable learning process is critically enabled by our second contribution: a robust distributional critic, which we design through a fusion of distributional RL and clipped double Q-learning. The resulting algorithm is highly effective, achieving state-of-the-art performance on a benchmark of eighteen hard RL tasks, including Humanoid, Dog, and Shadow Hand domains, spanning both dense-reward and goal-conditioned RL scenarios. Beyond standard benchmarks, we also evaluate a biologically motivated predator-prey task to examine the behavioral robustness and generalization capacity of our approach. Code: https://github.com/d2ac-actor-critic/d2ac-public
title D2 Actor Critic: Diffusion Actor Meets Distributional Critic
topic Machine Learning
url https://arxiv.org/abs/2510.03508