Beyond the Majority: Long-tail Imitation Learning for Robotic Manipulation

2026-02-06

AI Summary

This work addresses the challenge of limited generalization in universal robotic policies trained on long-tailed imitation learning data, where tasks in the tail suffer from severe sample scarcity. The authors demonstrate that conventional resampling strategies are ineffective in this setting and that data scarcity specifically impairs the policy’s spatial reasoning capabilities. To mitigate this, they propose Approaching-Phase Augmentation (APA), a novel knowledge transfer method that requires no external demonstrations. APA explicitly models task phases and enhances knowledge transfer from head to tail tasks by augmenting the approaching phase of manipulation trajectories. Experiments in both simulated and real-world manipulation tasks validate the effectiveness of APA, showing significant improvements in policy performance on tail tasks without compromising overall accuracy.

Technology Category

Intelligent Robots: ManipulationMachine Learning: Imitation Learning & Inverse Reinforcement LearningHumans and AI: Human-Aware Planning and Behavior Prediction

Application Category

Responsible Web: Machine-in-the-loop, human agency and autonomySearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingUser Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

While generalist robot policies hold significant promise for learning diverse manipulation skills through imitation, their performance is often hindered by the long-tail distribution of training demonstrations. Policies learned on such data, which is heavily skewed towards a few data-rich head tasks, frequently exhibit poor generalization when confronted with the vast number of data-scarce tail tasks. In this work, we conduct a comprehensive analysis of the pervasive long-tail challenge inherent in policy learning. Our analysis begins by demonstrating the inefficacy of conventional long-tail learning strategies (e.g., re-sampling) for improving the policy's performance on tail tasks. We then uncover the underlying mechanism for this failure, revealing that data scarcity on tail tasks directly impairs the policy's spatial reasoning capability. To overcome this, we introduce Approaching-Phase Augmentation (APA), a simple yet effective scheme that transfers knowledge from data-rich head tasks to data-scarce tail tasks without requiring external demonstrations. Extensive experiments in both simulation and real-world manipulation tasks demonstrate the effectiveness of APA. Our code and demos are publicly available at: https://mldxy.github.io/Project-VLA-long-tail/.

Problem

• long-tail distribution
• imitation learning
• robotic manipulation
• data scarcity
• generalization

Innovation

• long-tail learning
• imitation learning
• robotic manipulation
• spatial reasoning
• Approaching-Phase Augmentation