Galbot · robot AI model
StereoVLA
China · 2025
Sources retrieved: 18 November 2026 (A, B, C, D, E)
- Model type
- Vision-language-action (VLA)Source and note: Model typeD shengliangd.github.io/StereoVLA-Webpage/ · Retrieved 18 November 2026Note. Developer wording: “we present StereoVLA, a VLA model that leverages rich geometric cues from stereo vision”.
- What it controls
- Arms and handsSource and note: What it controlsB arxiv.org/html/2512.21970 · Retrieved 18 November 2026Note. Technical report: the action expert “predicts action chunks via flow-matching” with a delta end-effector pose representation; “the Franka Panda remains the primary experimental platform”.
- Inputs
- Camera LanguageSource and note: InputsA arxiv.org/abs/2512.21970 · Retrieved 18 November 2026Note. Paper abstract: the vision encoder “extracts geometric cues from subtle stereo-view disparities” and captures semantic features “to support language-conditioned manipulation”. Input is a stereo camera pair.
- Access
- Open weights (download)Source and note: AccessC huggingface.co/api/models/shengliangd/StereoVLA · Retrieved 18 November 2026Note. Hugging Face repository file list: “stereovla/checkpoint/model.safetensors”.
- Base model
- InternLM-1.8BSource and note: Base modelB arxiv.org/html/2512.21970 · Retrieved 18 November 2026Note. Technical report: the visual and language tokens are passed to a “pre-trained large language model, InternLM-1.8B”.
- Licence type
- Free, research onlySource and note: Licence typeE huggingface.co/shengliangd/StereoVLA · Retrieved 18 November 2026Note. Hugging Face model page: “License: cc-by-nc-4.0”; the model card is empty.

Specifications
8 of 14 stated 6 not stated
3 fields hidden 3 not-stated fields with a note stay ·
How to read the figuresShow guideClose
- not statedNo source
- 0Reported as zero
- ASource
Select a source marker for source details and notes.
Release1/1 disclosed
Input and output3/4 disclosed
Size and speed0/2 disclosed
Where it runs and what it is built on1/2 disclosed
Openness3/3 disclosed
The developer’s own claims0/2 disclosed
Notes
The project page and the Hugging Face weights are published under the first author's personal accounts (shengliangd). The project page lists the affiliations in the order “Galbot”, then “Peking University”, among others; the first author is affiliated with Galbot.
Model weights are published on Hugging Face under CC BY-NC 4.0: the repository lists “stereovla/checkpoint/model.safetensors”.
Sources
One line per source, not per figure: a source has one retrieval date.
- Aarxiv.org/abs/2512.21970Retrieved 18 November 2026
- Barxiv.org/html/2512.21970Retrieved 18 November 2026
- Chuggingface.co/api/models/shengliangd/StereoVLARetrieved 18 November 2026
- Dshengliangd.github.io/StereoVLA-Webpage/Main source. Retrieved 18 November 2026
- Ehuggingface.co/shengliangd/StereoVLARetrieved 18 November 2026





