الملخص
Training vision-language-action (VLA) models for high-precision manipulation typically requires task-specific, high-quality data (e.g., teleoperation), which is slow and expensive to collect. To reduce this burden without compromising manipulation precision, we propose $\varepsilon$4P (Imperfection for Precision), a simple yet effective method that "upcycles" two otherwise discarded data sources: (1) low-precision data from the target task and (2) high-precision data from mismatched tasks. Rather than naively mixing these imperfect data sources throughout co-training, $\varepsilon$4P controls where each source contributes along the flow-matching trajectory. Specifically, low-precision, target-task data is used at high noise to preserve high-level task context and high-precision, task-mismatched data is used at low noise to transfer low-level action precision. Through real-robot experiments on both sub-millimeter, high-precision tasks and coarse-grained tasks, we demonstrate that the proposed method (1) effectively leverages additional imperfect data to improve policy performance by up to 31.7 percentage points, and (2) can replace an equal amount of task-specific, high-quality data with an average performance drop of only 4.2 percentage points. Overall, $\varepsilon$4P points toward a scalable paradigm for high-precision manipulation, in which heterogeneous, imperfect data can be systematically repurposed to reduce reliance on costly task-specific, high-quality data. More details are available at https://varepsilon4p.github.io/.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Wei, H., Liu, Y., Tang, C., Li, S., Chen, J., Zhu, J., Wang, J., Yin, H., Cao, Z., & Li, T. (2026). Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation. https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation
MLA 9
Wei, Hao, et al. "Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation." https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation.
شيكاغو (المؤلف–التاريخ)
Wei, Hao, Yang Liu, Chao Tang, Shengbao Li, Jiangtao Chen, Jinxuan Zhu, Jiaheng Wang, Hong Yin, Zhaofeng Cao, and Tingguang Li. 2026. "Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation." https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation.
هارفارد
Wei, H., Liu, Y., Tang, C., Li, S., Chen, J., Zhu, J., Wang, J., Yin, H., Cao, Z. and Li, T. (2026) 'Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation', Available at: https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation.
فانكوفر
Wei H, Liu Y, Tang C, Li S, Chen J, Zhu J, et al. Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation. https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation
IEEE
H. Wei, Y. Liu, C. Tang, S. Li, J. Chen, J. Zhu, J. Wang, H. Yin, Z. Cao, and T. Li, "Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation," https://omanscience.com/ar/articles/imperfection-for-precision-upcycling-imperfect-data-for-high-precision-robotic-manipulation.