RoboRefer
- AI model
- Humanoids and robot learning
A vision-language model that locates the objects and free spaces a natural-language prompt refers to, with step-by-step spatial reasoning, for robotics use.
Source on GitHub: github.com/Zhoues/RoboRefer (opens in a new tab). Facts read at commit d97a995 (opens in a new tab).
Licenses
- Code
- Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit d97a995 (opens in a new tab), on 16 Oct 2026.
- Weights
- Apache-2.0 (opens in a new tab)
Read on huggingface.co/Zhoues/RoboRefer-2B-SFT (opens in a new tab).
- Commercial use
- Yes
What it does
- Resolves spatial referring prompts, such as pointing to a specific object or the free area between objects, in tabletop and outdoor scenes.
- Accepts RGB input alone or RGB-D input, and can estimate depth from RGB using Depth-Anything-V2 when depth data is missing.
- Provides model weights at 2B and 8B sizes, served through an inference API server that clients query with images and prompts.
- Releases the RefSpatial training dataset and SFT training scripts, with depth alignment and fine-tuning steps.
- Includes evaluation code for the RefSpatial-Bench and RefSpatial-Expand-Bench benchmarks.
Written by Robotheca from the project's README (opens in a new tab) at commit d97a995.
Getting started
- Difficulty
- Advanced
- How to start
- The project's website (opens in a new tab)
Maintenance
- Status
- Maintained: a commit in the last 12 months
- Newest commit
- 16 Dec 2025 on main
- Latest release
- None published on GitHub
Checked on GitHub on 16 Oct 2026.
Links
Robots on Robotheca that use it
No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.
Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.