Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and Fragmentation

TL;DR

This paper investigates how dataset diversity and fragmentation contribute to shortcut learning in generalist robot policies, revealing that limited diversity and distributional disparities hinder generalization, and proposes data augmentation strategies to mitigate this issue.

Contribution

It identifies dataset fragmentation and limited diversity as key causes of shortcut learning and offers data augmentation methods to improve policy generalization.

Findings

Dataset fragmentation leads to reliance on task-irrelevant features.

Data augmentation reduces shortcut learning and improves generalization.

Strategies are effective in both simulation and real-world environments.

Abstract

Generalist robot policies trained on large-scale datasets such as Open X-Embodiment (OXE) demonstrate strong performance across a wide range of tasks. However, they often struggle to generalize beyond the distribution of their training data. In this paper, we investigate the underlying cause of this limited generalization capability. We identify shortcut learning -- the reliance on task-irrelevant features -- as a key impediment to generalization. Through comprehensive theoretical and empirical analysis, we uncover two primary contributors to shortcut learning: (1) limited diversity within individual sub-datasets, and (2) significant distributional disparities across sub-datasets, leading to dataset fragmentation. These issues arise from the inherent structure of large-scale datasets like OXE, which are typically composed of multiple sub-datasets collected independently across varied environments and embodiments. Our findings provide critical insights into dataset collection strategies that can reduce shortcut learning and enhance the generalization ability of generalist robot policies. Moreover, in scenarios where acquiring new large-scale data is impractical, we demonstrate that carefully selected robotic data augmentation strategies can effectively reduce shortcut learning in existing offline datasets, thereby improving generalization capabilities of generalist robot policies, e.g., $\pi_0$, in both simulation and real-world environments. More information at https://lucky-light-sun.github.io/proj/shortcut-learning-in-grps/.