--- title: GEM-2B canonical_url: "https://www.modelscope.cn/models/AI-ModelScope/GEM-2B" md_url: "https://www.modelscope.cn/models/AI-ModelScope/GEM-2B.md" repository: AI-ModelScope/GEM-2B last_updated: 2026-06-01 license: mit pipeline_tag: image-text-to-text tasks: - image-text-to-text model_type: - qwen3_vl architectures: - Qwen3VL_Latent_ForConditionalGeneration base_model: - Qwen/Qwen3-VL-2B-Instruct base_model_relation: finetune parameters: 4.4B tensor_type: - BF16 library_name: - safetensors - pytorch frameworks: - pytorch language: - en - zh downloads: 40 stars: 0 tags: - Vision-Language-Model - Vision-Language-Action --- # GEM-2B > GEM-2B - AI-ModelScope 在 ModelScope 开源的模型。GEM: Generative Supervision Helps Embodied Intelligence AI-ModelScope/GEM-2B 是 ModelScope 魔搭社区上的 4.4B 参数image-text-to-text模型，采用 mit 许可，基于 Qwen/Qwen3-VL-2B-Instruct 构建。 - **Repository**: AI-ModelScope/GEM-2B - **License**: mit - **Tasks**: image-text-to-text - **Parameters**: 4.4B - **Base model**: Qwen/Qwen3-VL-2B-Instruct - **Tags**: Vision-Language-Model, Vision-Language-Action - **Downloads**: 40 - **Stars**: 0 - **Last updated**: 2026-06-01 Source: https://www.modelscope.cn/models/AI-ModelScope/GEM-2B ---

GEM: Generative Supervision Helps Embodied Intelligence

Ruowen Zhao¹, Bangguo Li¹, Zuyan Liu^1,2,†, Yinan Liang¹, Junliang Ye¹, Fangfu Liu¹,
Diankun Wu¹, Zhengyi Wang¹, Xumin Yu², Yongming Rao^2,✉, Han Hu², Jun Zhu^1,✉
^†Project Lead.^✉Corresponding Author.
¹Tsinghua University, ²Tencent Hunyuan

Embodied Vision-Language Models (VLMs) have demonstrated impressive performance and generalization in robotics, particularly within Vision-Language-Action frameworks. However, a significant gap remains between the high-level semantic focus of standard text-guided pre-training paradigms and the low-level spatial and physical knowledge critical for execution in embodied environments. In this paper, we introduce **GEM**, a Generative-supervised Embodied vision-language Model designed to bridge this divide. We propose integrating a depth map generation task directly into the VLM pre-training phase. By training this generative objective jointly with the main model, we observe substantial improvements in embodied intelligence, significantly enhancing both semantic understanding and physical operation capabilities. To support this paradigm, we curate and release GEM-4M, a comprehensive large-scale dataset featuring a mixture of grounding, reasoning, and planning data paired with high-quality depth supervision. Extensive experiments demonstrate that GEM achieves state-of-the-art results across diverse embodied benchmarks. Furthermore, our deployed action model, GEM-VLA, exhibits vastly superior task execution abilities in both simulation environments and real-world evaluations.

If you find our work helpful, please consider citing: ``` @article{zhao2026gem, title={GEM: Generative Supervision Helps Embodied Intelligence}, author={Zhao, Ruowen and Li, Bangguo and Liu, Zuyan and Liang, Yinan and Ye, Junliang and Liu, Fangfu and Wu, Diankun and Wang, Zhengyi and Yu, Xumin and Rao, Yongming and others}, journal={arXiv preprint arXiv:2605.28548}, year={2026} } ```