RESEARCH PAPER

GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation

Yupeng Zheng; Xiang Li; Songen Gu; Yuhang Zheng; Shuai Tian; Weize Li; Linbo Wang; Chaoyue Li; Qichao Zhang; Haoran Li; Zhongpu Xia; Ya-Qin Zhang; Shuicheng Yan; Dongbin Zhao

Classification

View four quadrants
Major category
WAMs
Quadrant
Not assigned
Architecture
Not assigned
Prediction paradigm
Not assigned
Source review status
Not assigned

Category review. The paper contributes task-specific structural supervision and evaluates complete VLA and WAM variants, including a future-generating model with inverse-dynamics action extraction. A WAM methods placement is supported, while no single architecture quadrant describes every variant. Reading evidence

AT A GLANCE

Contribution

GIFT trains robot-policy features to retain geometry, object–end-effector relations and instruction-relevant regions. The same auxiliary objectives improve three different action formulations while their default deployment uses no auxiliary predictions. Evidence is strongest for matched-baseline robustness gains, with substantial annotation requirements and uneven benefits across shifts.

Abstract

An abstract has not been added yet.

Affiliations

Institute of Automation, Chinese Academy of Sciences; University of Chinese Academy of Sciences; Tsinghua University; Fudan University; National University of Singapore