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

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

|Authors: 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

|Venue: arXiv preprint | Year: 2026

|arXiv: 2609.04193v1

Abstract

Vision-language pre-training and predictive world modeling provide robot policies with rich semantic and dynamic visual features, but their native action and visual-prediction objectives may omit critical physical and task structure while retaining control-irrelevant visual redundancy. We call this mismatch between visual richness and control utility the action-sufficiency gap. We investigate whether this gap can be bridged by guiding intermediate features to preserve three control-relevant structure in robotic manipulation: geometry governing motion feasibility, affordance encoding instruction-relevant entities, and goals grounding instructions in task-relevant regions. To this end, we present GIFT (Guided Intermediate Feature Training), an architecture-flexible framework for learning intermediate features that translates these structures into training-time constraints through geometry alignment, affordance prediction, and goal-region reconstruction. We instantiate GIFT in a Vision-Language-Action (VLA) policy, a direct-action World-Action Model (WAM), and an inverse-dynamics WAM while retaining each model’s action formulation. Under zero-shot transfer to LIBERO-Plus, GIFT-VLA, GIFT-WAM-Fast, and GIFT-WAM-IDM outperform StarVLA-OFT, Fast-WAM, and Fast-WAM-IDM by 4.6, 12.6, and 5.2 points, reaching 79.6%, 72.6%, and 87.8%, respectively. On RoboCasa, the three GIFT variants reach 61.4%, 83.6%, and 82.3%, outperforming their counterparts by 12.6, 9.0, and 8.4 points, respectively. Together, these results establish learning functionally structured intermediate features as a reusable principle across model-specific action formulations, with especially large gains on articulated-object tasks and high-precision real-world manipulation under unseen visual and spatial perturbations. Project page: https://openphoenix-team.github.io/GIFT-pages.

Key Contributions

• Vision-language pre-training and predictive world modeling provide robot policies with rich seman…
• We call this mismatch between visual richness and control utility the action-sufficiency gap
• We investigate whether this gap can be bridged by guiding intermediate features to preserve three…
• To this end, we present GIFT (Guided Intermediate Feature Training), an architecture-flexible fra…
• We instantiate GIFT in a Vision-Language-Action (VLA) policy, a direct-action World-Action Model …

Topics

• manipulation
• vision
• vla
• reinforcement-learning
• control

Code & Data

No code repository linked in paper metadata.

BibTeX

@article{Zheng2026_260904193v1,
  title     = {GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation},
  author    = {Yupeng Zheng and Xiang Li and Songen Gu and Yuhang Zheng and Shuai Tian and Weize Li and Linbo Wang and Chaoyue Li and Qichao Zhang and Haoran Li and Zhongpu Xia and Ya-Qin Zhang and Shuicheng Yan and Dongbin Zhao},
  year      = {2026},
  eprint    = {2609.04193v1},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url       = {https://arxiv.org/abs/2609.04193v1}
}