hugging-apps/embodied-r1-5

Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models

This Space demos Embodied-R1.5, an 8B parameter embodied foundation model built on Qwen3-VL-8B-Instruct. It unifies spatial cognition & reasoning, task planning & correction, and embodied pointing & location within a single architecture.

Usage

1. 1.Upload a scene image
2. 2.Enter a task instruction or question about the scene
3. 3.Click "Run Inference"

The model reasons about the visual observation and emits its final answer inside an <answer> tag. If the answer contains 2D points (normalized to [0, 1000]), they are automatically visualized on the image.

Model

• —Model: IffYuan/Embodied-R1.5
• —Architecture: Qwen3-VL-8B-Instruct (fine-tuned)
• —Paper: Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models
• —Code: GitHub

Supported Task Types

The model supports spatial reasoning, task planning, pointing/grounding, and open-ended VQA. All points and boxes are normalized to [0, 1000] regardless of the original image resolution.