DiWA: Diffusion Policy Adaptation with World Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chandra, Akshay L, Nematollahi, Iman, Huang, Chenguang, Welschehold, Tim, Burgard, Wolfram, Valada, Abhinav
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912521639690240
author Chandra, Akshay L
Nematollahi, Iman
Huang, Chenguang
Welschehold, Tim
Burgard, Wolfram
Valada, Abhinav
author_facet Chandra, Akshay L
Nematollahi, Iman
Huang, Chenguang
Welschehold, Tim
Burgard, Wolfram
Valada, Abhinav
contents Fine-tuning diffusion policies with reinforcement learning (RL) presents significant challenges. The long denoising sequence for each action prediction impedes effective reward propagation. Moreover, standard RL methods require millions of real-world interactions, posing a major bottleneck for practical fine-tuning. Although prior work frames the denoising process in diffusion policies as a Markov Decision Process to enable RL-based updates, its strong dependence on environment interaction remains highly inefficient. To bridge this gap, we introduce DiWA, a novel framework that leverages a world model for fine-tuning diffusion-based robotic skills entirely offline with reinforcement learning. Unlike model-free approaches that require millions of environment interactions to fine-tune a repertoire of robot skills, DiWA achieves effective adaptation using a world model trained once on a few hundred thousand offline play interactions. This results in dramatically improved sample efficiency, making the approach significantly more practical and safer for real-world robot learning. On the challenging CALVIN benchmark, DiWA improves performance across eight tasks using only offline adaptation, while requiring orders of magnitude fewer physical interactions than model-free baselines. To our knowledge, this is the first demonstration of fine-tuning diffusion policies for real-world robotic skills using an offline world model. We make the code publicly available at https://diwa.cs.uni-freiburg.de.
format Preprint
id arxiv_https___arxiv_org_abs_2508_03645
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DiWA: Diffusion Policy Adaptation with World Models
Chandra, Akshay L
Nematollahi, Iman
Huang, Chenguang
Welschehold, Tim
Burgard, Wolfram
Valada, Abhinav
Robotics
Computer Vision and Pattern Recognition
Machine Learning
Fine-tuning diffusion policies with reinforcement learning (RL) presents significant challenges. The long denoising sequence for each action prediction impedes effective reward propagation. Moreover, standard RL methods require millions of real-world interactions, posing a major bottleneck for practical fine-tuning. Although prior work frames the denoising process in diffusion policies as a Markov Decision Process to enable RL-based updates, its strong dependence on environment interaction remains highly inefficient. To bridge this gap, we introduce DiWA, a novel framework that leverages a world model for fine-tuning diffusion-based robotic skills entirely offline with reinforcement learning. Unlike model-free approaches that require millions of environment interactions to fine-tune a repertoire of robot skills, DiWA achieves effective adaptation using a world model trained once on a few hundred thousand offline play interactions. This results in dramatically improved sample efficiency, making the approach significantly more practical and safer for real-world robot learning. On the challenging CALVIN benchmark, DiWA improves performance across eight tasks using only offline adaptation, while requiring orders of magnitude fewer physical interactions than model-free baselines. To our knowledge, this is the first demonstration of fine-tuning diffusion policies for real-world robotic skills using an offline world model. We make the code publicly available at https://diwa.cs.uni-freiburg.de.
title DiWA: Diffusion Policy Adaptation with World Models
topic Robotics
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2508.03645