UniFS: Unified Fast-to-Slow Hierarchical Architecture for Vision-Language-Action Models

📄 arXiv: 2606.22794v1 📥 PDF

作者: Lin Sun, Zhiwei Guan, Conglin Wang, Zihong Chen, Jianhai Yu, Zongsheng Li, Boyong He, Tao Sun, Jiale Cao, Lige Liu

分类: cs.RO

发布日期: 2026-06-22

备注: Code is opensourced at https://github.com/linsun449/UniFS

🔗 代码/项目: GITHUB

💡 一句话要点

提出UniFS以解决视觉-语言-动作模型中的频率困境

🎯 匹配领域: 支柱一：机器人控制 (Robot Control) 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 视觉-语言-动作 多模态学习 信息耦合 层次架构 机器人控制

📋 核心要点

  1. 现有的快慢双系统模型在频率更新上存在矛盾，导致信息传递不畅和语义漂移。
  2. UniFS通过分层更新、潜在向量反转和多级监督策略，优化了视觉-语言模型与动作专家之间的信息交互。
  3. 在LIBERO数据集上，UniFS取得了98.3%的成功率，且推理速度提升了2.1倍，验证了其有效性。

📝 摘要（中文）

主流的快慢双系统视觉-语言-动作模型将高频动作专家与低频视觉-语言模型解耦以提高效率，但面临频率困境：更新间隔过大导致语义漂移，过小则削弱计算节省。此外，动作专家仅接收视觉-语言模型最后一层的固定频率表示，导致丰富的中间特征被丢弃，限制了信息耦合和操作精度。为此，本文提出UniFS，一个统一的快慢层次架构，通过分层更新、潜在向量反转机制和多级监督策略来解决这些挑战。实验表明，UniFS在LIBERO数据集上实现了98.3%的平均成功率，较VLA-Adapter基线提升2.5%，并将平均推理延迟从36.5毫秒降低至17.8毫秒，验证了其在实际机器人平台上的应用潜力。

🔬 方法详解

问题定义：本文旨在解决现有快慢双系统视觉-语言-动作模型中的频率更新矛盾，导致的语义漂移和信息损失问题。现有方法在高频和低频之间的解耦使得信息传递不够高效。

核心思路：UniFS的核心思路是通过分层设计和多级监督策略，优化信息流动，确保快速变化的动态与稳定的语义上下文之间的有效耦合。

技术框架：UniFS的整体架构将视觉-语言模型的层次分为多个组，采用逐层递减的更新频率。浅层捕捉快速变化的动态，而深层则缓存稳定的语义信息。潜在向量反转机制重新调整多尺度特征与动作专家的交互顺序。

关键创新：UniFS的主要创新在于其统一的快慢层次架构，允许在单一骨干网络中实现丰富的跨频率信息传递，解决了传统方法中信息流失的问题。

关键设计：在设计中，采用了分层更新机制、潜在向量反转机制和多级监督策略，确保了信息的有效耦合与传递，同时保留了时间上下文。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

UniFS在LIBERO数据集上实现了98.3%的平均成功率，较VLA-Adapter基线提升了2.5%。同时，推理延迟从36.5毫秒降低至17.8毫秒，实现了2.1倍的速度提升，展现了其优越的性能和实用性。

🎯 应用场景

UniFS的研究成果在多模态学习、机器人控制和人机交互等领域具有广泛的应用潜力。其高效的信息处理能力能够提升智能系统在复杂环境中的决策和执行能力，推动智能机器人在实际任务中的应用。

📄 摘要（原文）

Mainstream Fast-Slow dual system vision-language-action models decouple a high-frequency action expert from a low-frequency vision-language model for efficiency, yet they face a fundamental frequency dilemma: large update gaps cause semantic drift from stale context, while small gaps erode the intended computational savings. Moreover, because the action expert receives only the VLM's final-layer representation at a single fixed frequency, rich intermediate features are discarded, limiting both information coupling and manipulation precision. Inspired by multi-timescale neural processing in the human brain, we introduce UniFS, a unified fast-to-slow architecture that resolves these challenges through three key designs. First, we stratify the VLM layers into groups with progressively decreasing update frequencies, enabling shallow layers to capture fast-changing dynamics while deeper layers cache stable semantic context. Second, a latent vector inversion mechanism re-routes the interaction order between multi-scale VLM features and the action expert, aligning fast-varying representations with fine-grained action decoding and slow-varying ones with coarse planning. Third, a multi-level supervision strategy enforces a coarse-to-fine learning hierarchy across temporal scales. Together, these designs enable richer cross-frequency information transfer within a single backbone, while the low-frequency pathways additionally preserve temporal context across steps. Experiments on LIBERO show that UniFS achieves state-of-the-art performance (98.3\% average success rate, a 2.5\% gain over VLA-Adapter baseline) while reducing average inference latency from 36.5~ms to 17.8~ms (2.1$\times$ speedup). Real-robot experiments on a Franka platform further validate its practical applicability. Code is opensourced at https://github.com/linsun449/UniFS.