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

Hugging Face Daily Papers Papers

Summary

StreamPI introduces a streaming multimodal temporal modeling framework for vision-language-action models, improving robot manipulation through instruction-anchored attention and randomized interval training without additional parameters.

Vision-Language-Action (VLA) models have demonstrated effectiveness in robot manipulation, yet state-of-the-art models such as pi0.5 operate under a single-frame paradigm, limiting their ability to retain past observations and develop precise spatial perception. In this paper, we propose StreamPI, a streaming multimodal temporal modeling framework that equips single-frame VLA with temporal reasoning capability without introducing any additional parameters. One core design is instruction-anchored temporal modeling. It treats each (visual observation, language instruction) pair as an atomic temporal unit: bidirectional attention within each pair enables cross-modal fusion, while causal attention across pairs preserves autoregressive streaming inference. This ensures the language instruction serves as a persistent semantic anchor throughout task execution. To bridge the gap between synchronous training and asynchronous real-robot deployment, we introduce a andom-interval streaming training strategy: a proper inter-frame interval (e.g., every 3 frames) enables faster and smoother action execution. Beyond this, randomizing the interval further improves robustness to frame-timing perturbations, supporting asynchronous deployment in practice. Furthermore, by leveraging the length extrapolation capability of the LLM backbone, StreamPI seamlessly inherits pretrained single-frame weights and supports flexible single-frame and multi-frame inference. Experiments on real-robot tasks spanning memory-dependent and precise perception scenarios, as well as the simulation benchmark LIBERO, demonstrate that StreamPI outperforms pi0.5 across diverse tasks.
Original Article
View Cached Full Text

Cached at: 08/27/26, 07:20 AM

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

Source: https://huggingface.co/papers/2608.26067

Abstract

StreamPI enhances single-frame vision-language-action models with streaming temporal reasoning via instruction-anchored attention and randomized interval training, improving robot manipulation without extra parameters.

Vision-Language-Action (VLA) models have demonstrated effectiveness in robot manipulation, yet state-of-the-art models such as pi0.5 operate under a single-frame paradigm, limiting their ability to retain past observations and develop precise spatial perception. In this paper, we propose StreamPI, astreaming multimodal temporal modelingframework that equips single-frame VLA with temporal reasoning capability without introducing any additional parameters. One core design isinstruction-anchored temporal modeling. It treats each (visual observation, language instruction) pair as an atomic temporal unit:bidirectional attentionwithin each pair enables cross-modal fusion, whilecausal attentionacross pairs preservesautoregressive streaming inference. This ensures the language instruction serves as a persistent semantic anchor throughout task execution. To bridge the gap between synchronous training and asynchronous real-robot deployment, we introduce a andom-interval streaming training strategy: a proper inter-frame interval (e.g., every 3 frames) enables faster and smoother action execution. Beyond this, randomizing the interval further improves robustness to frame-timing perturbations, supporting asynchronous deployment in practice. Furthermore, by leveraging thelength extrapolationcapability of theLLM backbone, StreamPI seamlessly inherits pretrained single-frame weights and supports flexible single-frame and multi-frame inference. Experiments on real-robot tasks spanning memory-dependent and precise perception scenarios, as well as the simulation benchmark LIBERO, demonstrate that StreamPI outperforms pi0.5 across diverse tasks.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2608\.26067

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.26067 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.26067 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.26067 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

See like a Robot: Robot-Centric Pointmaps for Vision-Language-Action Models

Hugging Face Daily Papers

This paper introduces robot-centric pointmaps, which encode 3D scene coordinates in the robot frame directly into image pixels, to resolve the frame mismatch between camera observations and robot action definitions in Vision-Language-Action models. The method improves Pi0.5 and SmolVLA on RoboCasa and generalizes better to unseen camera placements in real-robot experiments.