---
title: "HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents"
canonical_url: "https://www.modelscope.cn/papers/265178"
md_url: "https://www.modelscope.cn/papers/265178.md"
arxiv_id: 2604.07430
published: 2026-04-08
last_updated: 2026-04-08
authors:
  - "Tencent Robotics X"
  - "HY Vision Team"
  - "Xumin Yu"
  - "Zuyan Liu"
  - "Ziyi Wang"
  - "He Zhang"
  - "Yongming Rao"
  - "Fangfu Liu"
  - "Yani Zhang"
  - "Ruowen Zhao"
  - "Oran Wang"
  - "Yves Liang"
  - "Haitao Lin"
  - "Minghui Wang"
  - "Yubo Dong"
  - "Kevin Cheng"
  - "Bolin Ni"
  - "Rui Huang"
  - "Han Hu"
  - "Zhengyou Zhang"
  - Linus
  - "Shunyu Yao"
model_name: HY-Embodied-0.5
model_developer: "Tencent Robotics X、Tencent Hunyuan Team"
domain:
  - "计算机视觉"
  - "自然语言处理"
  - "多模态大模型"
  - "具身智能"
  - "机器人学"
type:
  - "计算机视觉"
  - "自然语言处理"
  - "多模态大模型"
  - "具身智能"
  - "机器人学"
  - "Computer Vision and Pattern Recognition"
arxiv_url: "https://arxiv.org/abs/2604.07430"
pdf_url: "https://arxiv.org/pdf/2604.07430"
code_link: "https://github.com/Tencent-Hunyuan/HY-Embodied"
---

# HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents

> We introduce HY-Embodied-0.5, a family of foundation models specifically designed for real-world embodied agents. To bridge the gap between general Vision-Language Models (VLMs) and the demands of embodied agents, our models are developed to enhance the core…

「HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents」是 ModelScope 魔搭社区收录的论文，arXiv 2604.07430，作者为 Tencent Robotics X, HY Vision Team, Xumin Yu et al.，发表于 2026-04-08，属于 计算机视觉、自然语言处理、多模态大模型 领域。

- **ArXiv**: 2604.07430
- **Published**: 2026-04-08
- **Authors**: Tencent Robotics X, HY Vision Team, Xumin Yu, Zuyan Liu, Ziyi Wang, He Zhang, Yongming Rao, Fangfu Liu, Yani Zhang, Ruowen Zhao, Oran Wang, Yves Liang, Haitao Lin, Minghui Wang, Yubo Dong, Kevin Cheng, Bolin Ni, Rui Huang, Han Hu, Zhengyou Zhang, Linus, Shunyu Yao
- **Model**: HY-Embodied-0.5
- **Developer**: Tencent Robotics X、Tencent Hunyuan Team
- **Domain**: 计算机视觉, 自然语言处理, 多模态大模型, 具身智能, 机器人学
- **ArXiv URL**: https://arxiv.org/abs/2604.07430
- **PDF**: https://arxiv.org/pdf/2604.07430
- **Code**: https://github.com/Tencent-Hunyuan/HY-Embodied

Source: https://www.modelscope.cn/papers/265178

---

> HY-Embodied-0.5：打通数字智能与物理世界的具身基础模型双轨引擎

## 摘要

本文针对现实世界具身智能体（embodied agents）对细粒度视觉感知与物理交互推理的迫切需求，提出HY-Embodied-0.5系列具身基础模型。研究背景在于现有通用视觉语言模型（VLMs）虽在图文理解上表现优异，但在空间建模、时序感知、动作预测与物理环境规划等具身任务上存在显著短板。为此，作者从架构、数据与训练三方面系统创新：架构上采用轻量原生分辨率ViT（HY-ViT 2.0）、模态自适应的混合Transformer（MoT）结构，并引入视觉隐式标记（visual latent tokens）强化跨模态对齐；数据上构建超1亿高质量样本的多层级预训练语料，覆盖检测、深度估计、分割、指代计数、第一人称机器人操作及空间推理等具身特异性任务；训练上设计迭代式自进化后训练范式，结合强化学习与拒绝采样监督微调，并通过大模型到小模型的在线策略蒸馏（on-policy distillation）提升边缘部署版本性能。在22个视觉感知、空间推理与具身理解基准上，MoT-2B变体在16项中超越同规模SOTA，32B变体性能媲美Gemini 3.0 Pro；下游真实机器人控制实验中，基于该VLM训练的Vision-Language-Action（VLA）模型展现出强泛化性与实用性。

## Abstract

We introduce HY-Embodied-0.5, a family of foundation models specifically designed for real-world embodied agents. To bridge the gap between general Vision-Language Models (VLMs) and the demands of embodied agents, our models are developed to enhance the core capabilities required by embodied intelligence: spatial and temporal visual perception, alongside advanced embodied reasoning for prediction, interaction, and planning. The HY-Embodied-0.5 suite comprises two primary variants: an efficient model with 2B activated parameters designed for edge deployment, and a powerful model with 32B activated parameters targeted for complex reasoning. To support the fine-grained visual perception essential for embodied tasks, we adopt a Mixture-of-Transformers (MoT) architecture to enable modality-specific computing. By incorporating latent tokens, this design effectively enhances the perceptual representation of the models. To improve reasoning capabilities, we introduce an iterative, self-evolving post-training paradigm. Furthermore, we employ on-policy distillation to transfer the advanced capabilities of the large model to the smaller variant, thereby maximizing the performance potential of the compact model. Extensive evaluations across 22 benchmarks, spanning visual perception, spatial reasoning, and embodied understanding, demonstrate the effectiveness of our approach. Our MoT-2B model outperforms similarly sized state-of-the-art models on 16 benchmarks, while the 32B variant achieves performance comparable to frontier models such as Gemini 3.0 Pro. In downstream robot control experiments, we leverage our robust VLM foundation to train an effective Vision-Language-Action (VLA) model, achieving compelling results in real-world physical evaluations. Code and models are open-sourced at https://github.com/Tencent-Hunyuan/HY-Embodied.