Saved in:
Bibliographic Details
Main Authors: Tong, Zizhao, Chen, Di, Hu, Sicheng, Fan, Hongwei, Chen, Liliang, Ren, Guanghui, Tang, Hao, Dong, Hao, Shao, Ling
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.24797
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918150611664896
author Tong, Zizhao
Chen, Di
Hu, Sicheng
Fan, Hongwei
Chen, Liliang
Ren, Guanghui
Tang, Hao
Dong, Hao
Shao, Ling
author_facet Tong, Zizhao
Chen, Di
Hu, Sicheng
Fan, Hongwei
Chen, Liliang
Ren, Guanghui
Tang, Hao
Dong, Hao
Shao, Ling
contents Generalist robot policies trained on large-scale, visually homogeneous datasets can be susceptible to shortcut learning, which impairs their out-of-distribution (OOD) generalization. While generative data augmentation is a common approach to introduce diversity, it presents a subtle challenge: data composition. Naively mixing real and synthetic data can corrupt the learning signal, as this process often prioritizes visual diversity at the expense of information fidelity. This paper suggests that robust generalization depends on principled, fidelity-aware data composition. We introduce Coherent Information Fidelity Tuning (CIFT), a framework that treats data composition as an optimization problem. CIFT uses a practical proxy for Information Fidelity based on the feature-space geometry of a dataset. This enables the identification of a phase transition, termed the Decoherence Point, where training stability degrades. The framework includes a generative engine, Multi-View Video Augmentation (MVAug), to synthesize a causally disentangled data spectrum for this tuning process. Applying CIFT to policy architectures such as $π_0$ and Diffusion Policy improves OOD success rates by over 54\%. These results indicate that fidelity-aware composition, beyond data synthesis alone, is an important component for developing robust, general-purpose robots.
format Preprint
id arxiv_https___arxiv_org_abs_2509_24797
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fidelity-Aware Data Composition for Robust Robot Generalization
Tong, Zizhao
Chen, Di
Hu, Sicheng
Fan, Hongwei
Chen, Liliang
Ren, Guanghui
Tang, Hao
Dong, Hao
Shao, Ling
Robotics
Artificial Intelligence
Machine Learning
Generalist robot policies trained on large-scale, visually homogeneous datasets can be susceptible to shortcut learning, which impairs their out-of-distribution (OOD) generalization. While generative data augmentation is a common approach to introduce diversity, it presents a subtle challenge: data composition. Naively mixing real and synthetic data can corrupt the learning signal, as this process often prioritizes visual diversity at the expense of information fidelity. This paper suggests that robust generalization depends on principled, fidelity-aware data composition. We introduce Coherent Information Fidelity Tuning (CIFT), a framework that treats data composition as an optimization problem. CIFT uses a practical proxy for Information Fidelity based on the feature-space geometry of a dataset. This enables the identification of a phase transition, termed the Decoherence Point, where training stability degrades. The framework includes a generative engine, Multi-View Video Augmentation (MVAug), to synthesize a causally disentangled data spectrum for this tuning process. Applying CIFT to policy architectures such as $π_0$ and Diffusion Policy improves OOD success rates by over 54\%. These results indicate that fidelity-aware composition, beyond data synthesis alone, is an important component for developing robust, general-purpose robots.
title Fidelity-Aware Data Composition for Robust Robot Generalization
topic Robotics
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2509.24797