Towards Generalizable Robotic Data Flywheel: High-Dimensional Factorization and Composition

2026-03-26

AI Summary

General-purpose robotic models often suffer from limited generalization due to insufficient data diversity and inefficiency in leveraging task-relevant information. This work proposes the F-ACIL framework, which introduces, for the first time, a heuristic factor-aware compositional iterative learning mechanism. By decoupling demonstration data into structured factor spaces—such as objects, actions, and environments—and employing factor-level data collection with iterative training, F-ACIL efficiently exploits implicit high-dimensional task factors. The approach systematically organizes real-world robotic data to construct a scalable and generalizable data flywheel. In physical experiments, F-ACIL achieves over a 45% performance improvement while using 5–10 times fewer demonstrations than baseline methods, substantially enhancing compositional generalization capabilities.

Technology Category

Intelligent Robots: Learning & Optimization for ROBMachine Learning: Imitation Learning & Inverse Reinforcement LearningSearch and Optimization: Learning to Search

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

The lack of sufficiently diverse data, coupled with limited data efficiency, remains a major bottleneck for generalist robotic models, yet systematic strategies for collecting and curating such data are not fully explored. Task diversity arises from implicit factors that are sparsely distributed across multiple dimensions and are difficult to define explicitly. To address this challenge, we propose F-ACIL, a heuristic factor-aware compositional iterative learning framework that enables structured data factorization and promotes compositional generalization. F-ACIL decomposes the data distribution into structured factor spaces such as object, action, and environment. Based on the factorized formulation, we develop a factor-wise data collection and an iterative training paradigm that promotes compositional generalization over the high-dimensional factor space, leading to more effective utilization of real-world robotic demonstrations. With extensive real-world experiments, we show that F-ACIL can achieve more than 45% performance gains with 5-10$\times$ fewer demonstrations comparing to that of which without the strategy. The results suggest that structured factorization offers a practical pathway toward efficient compositional generalization in real-world robotic learning. We believe F-ACIL can inspire more systematic research on building generalizable robotic data flywheel strategies. More demonstrations can be found at: https://f-acil.github.io/

Problem

• data diversity
• data efficiency
• compositional generalization
• robotic learning
• factorization

Innovation

• factor-aware compositional learning
• structured data factorization
• compositional generalization
• robotic data flywheel
• iterative training paradigm