---
title: "Patch Policy：通过密集视觉表示实现高效具身控制"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.18236v1"
permalink: "https://feed.astralor.com/items/itm_71c8c53bf0caeaa21551a12b"
published: "2026-07-20T17:59:41.000Z"
score: 79
channel: "ai"
tags: ["OpenVLA","Patch Policy","Vision Transformer","具身智能","机器人学习","注意力机制"]
---

# Patch Policy：通过密集视觉表示实现高效具身控制

## AI 点评

> 这篇论文让我眼前一亮：它巧妙地以极低成本实现了密集视觉特征的接入，对资源受限的机器人团队尤为友好。相比于依赖数十亿参数VLM的VLA策略，Patch Policy的轻量设计可能成为未来高频控制的首选范式。如果你在从事机器人操作或视觉策略研究，强烈建议阅读全文。

## 速览

Patch Policy是一种轻量级Transformer策略架构，直接利用预训练ViT的密集patch特征，无需完整大视觉语言模型。通过块因果注意力掩码，在保持时间因果性的同时，让模型关注每个观测的多个patch token及状态信息。在4个仿真和3个真实环境套件中，相比全局池化策略提升40%相对性能，超越微调OpenVLA-OFT约18%，而参数量仅为后者的0.7%。该方法为机器人社区提供高效利用视觉预训练的管道。

## 关键点

- 提出Patch Policy架构，直接消费密集预训练patch token，避免完整VLM的计算开销。
- 核心是块因果注意力掩码，平衡时间因果性与空间密集注意力。
- 在7个环境套件上平均提升40%，并超越OpenVLA-OFT 18%而参数量仅为0.7%。
- 为机器人社区提供轻量级、高速的视觉预训练利用管道。

## 全文

Pretrained dense visual features from Vision Transformers (ViTs) are powerful yet have been underutilized in robot learning. Modern robot policies either compress each observation into a single global token, or rely on visual backbones trained from scratch, sacrificing both fine-grained spatial detail and the benefits of large-scale visual pre-training. While there exist policies that do operate on dense patch features like large vision-language-action models (VLAs), they tend to be heavy and slow, inheriting the full cost of a billion-parameter vision-language model (VLM) backbone. We close this gap with Patch Policy, a minimal architectural extension that enables transformer-based policies to consume dense pre-trained patch tokens directly without the computational overhead of a full VLM. At its core is a block-causal attention mask that preserves the temporal causality of standard policies while letting the model attend over many patch tokens per observation, alongside other state information. Patch Policy is lightweight, fast, and highly effective. Across four simulated and three real-world environment suites, our method achieves a 40% relative improvement over policies using state-of-the-art global-pooled representations. Furthermore, it surpasses fine-tuned OpenVLA-OFT by 18% while using roughly 0.7% of the parameters. We believe Patch Policy provides a pipeline for the robotics community to readily leverage continuing progress in visual representation learning, without sacrificing the training efficiency or inference speed required for high-frequency, reactive control. Videos can be viewed at https://patch-policy.github.io

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_71c8c53bf0caeaa21551a12b