---
title: "RoboTTT: Context Scaling for Robot Policies via Test-Time Training"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.15275v1"
permalink: "https://feed.astralor.com/items/itm_745c16e5b4b25039b35424fd"
published: "2026-07-16T17:59:06.000Z"
score: 79
channel: "ai"
tags: ["NVIDIA","RoboTTT","Vision-Language-Action","Context Scaling","Test-Time Training"]
---

# RoboTTT: Context Scaling for Robot Policies via Test-Time Training

## AI commentary

> This paper is worth watching because it demonstrates that 'context scaling' — a law similar to that in large language models — also applies to robot foundation models. For robotics researchers and practitioners, RoboTTT offers a path to improve performance via longer contexts without retraining the entire model. Particularly for companies working on long-horizon task planning and manipulation, this technique could significantly shorten deployment time.

## At a glance

The paper introduces RoboTTT, a novel method that scales visuomotor context for robot policies to 8K timesteps—three orders of magnitude beyond state-of-the-art—without increasing inference latency. By integrating Test-Time Training into Vision-Language-Action foundation models, RoboTTT enables one-shot in-context imitation from human video demonstrations, on-the-fly policy improvement, robustness to perturbations, and superior performance on multi-stage long-horizon tasks. On real-robot manipulation tasks, it achieves an 87% improvement over single-step baselines and completes a 5-minute, 10-stage assembly task that no baseline could. Pretraining with 8K-step context yields 62% better performance than 1K-step, suggesting context length as a new scaling axis for robot foundation models.

## Key points

- RoboTTT scales visuomotor context to 8K timesteps, three orders of magnitude beyond SOTA policies.
- Integrates Test-Time Training to compress and retrieve long contexts via gradient updates without increased latency.
- Achieves 87% improvement on real robot tasks and completes a 5-minute, 10-stage assembly task that no baseline does.
- Demonstrates context length as a new scaling axis for robot foundation models (62% gain from 8K over 1K pretraining).

## Full text

Recent robot foundation models operate with single-step or short-history visuomotor context. We introduce Test-Time-Training Robot Policies (RoboTTT), a robot model and training recipe that scale visuomotor context to 8K timesteps, three orders of magnitude beyond state-of-the-art policies, without growing inference latency. At this context length, we unlock new robot capabilities: one-shot in-context imitation from human video demonstrations, on-the-fly policy improvement, robustness to perturbations, and stronger performance on multi-stage, long-horizon tasks. We also observe, for the first time, steady gains in closed-loop performance as pretraining context length scales. At its core, RoboTTT integrates Test-Time Training into robot foundation models such as Vision-Language-Action policies, yielding a sequence model whose recurrent state consists of fast weights, parameters updated by gradient descent during both training and inference, compressing histories into weight space and retrieving contextual information for long-context conditioning. To scale training context length, the recipe combines sequence action forcing with truncated backpropagation through time. On challenging real-robot manipulation tasks, RoboTTT improves overall performance by 87% over the single-step context baseline and fully completes a five-minute, ten-stage assembly task, which no baseline ever does. RoboTTT trained with 8K-timestep context outperforms the same model pretrained with 1K timesteps by 62%, suggesting context length as a new scaling axis for robot foundation models. Videos are available at https://research.nvidia.com/labs/gear/robottt/

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_745c16e5b4b25039b35424fd