paper/arxiv/2608.26067

Robotics Dataset

StreamPI is a streaming multimodal temporal modeling framework that enhances single-frame Vision-Language-Action (VLA) models with temporal reasoning without adding parameters, using instruction-anchored temporal modeling and random-interval streaming training for robust asynchronous deployment.

Catalog metadata

Repository owner

Zhe Liu, Jinghua Hou, Yuxiang Lu et al.

Source

Research paper

Task

StreamPI: Streaming Multimodal Temporal Modeling for Vision-Language-Action Models

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.