Provable Ordering and Continuity in Vision-Language Pretraining for Generalizable Embodied Agents

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Zhizhen, Zhu, Lei, Fang, Zhen, Huang, Zi, Luo, Yadan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915682553167872
author Zhang, Zhizhen
Zhu, Lei
Fang, Zhen
Huang, Zi
Luo, Yadan
author_facet Zhang, Zhizhen
Zhu, Lei
Fang, Zhen
Huang, Zi
Luo, Yadan
contents Pre-training vision-language representations on human action videos has emerged as a promising approach to reduce reliance on large-scale expert demonstrations for training embodied agents. However, prior methods often employ time contrastive learning based on goal-reaching heuristics, progressively aligning language instructions from the initial to the final frame. This overemphasis on future frames can result in erroneous vision-language associations, as actions may terminate early or include irrelevant moments in the end. To address this issue, we propose Action Temporal Coherence Learning (AcTOL) to learn ordered and continuous vision-language representations without rigid goal-based constraint. AcTOL treats a video as a continuous trajectory where it (1) contrasts semantic differences between frames to reflect their natural ordering, and (2) imposes a local Brownian bridge constraint to ensure smooth transitions across intermediate frames. Extensive imitation learning experiments on both simulated and real robots show that the pretrained features significantly enhance downstream manipulation tasks with high robustness to different linguistic styles of instructions, offering a viable pathway toward generalized embodied agents.
format Preprint
id arxiv_https___arxiv_org_abs_2502_01218
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Provable Ordering and Continuity in Vision-Language Pretraining for Generalizable Embodied Agents
Zhang, Zhizhen
Zhu, Lei
Fang, Zhen
Huang, Zi
Luo, Yadan
Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
Pre-training vision-language representations on human action videos has emerged as a promising approach to reduce reliance on large-scale expert demonstrations for training embodied agents. However, prior methods often employ time contrastive learning based on goal-reaching heuristics, progressively aligning language instructions from the initial to the final frame. This overemphasis on future frames can result in erroneous vision-language associations, as actions may terminate early or include irrelevant moments in the end. To address this issue, we propose Action Temporal Coherence Learning (AcTOL) to learn ordered and continuous vision-language representations without rigid goal-based constraint. AcTOL treats a video as a continuous trajectory where it (1) contrasts semantic differences between frames to reflect their natural ordering, and (2) imposes a local Brownian bridge constraint to ensure smooth transitions across intermediate frames. Extensive imitation learning experiments on both simulated and real robots show that the pretrained features significantly enhance downstream manipulation tasks with high robustness to different linguistic styles of instructions, offering a viable pathway toward generalized embodied agents.
title Provable Ordering and Continuity in Vision-Language Pretraining for Generalizable Embodied Agents
topic Robotics
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2502.01218