Peking University · robot AI model
ViGAR
China · 2026
Sources retrieved: 19 November 2026 (A, B, C)
- Model type
- World action modelSource and note: Model typeA arxiv.org/abs/2610.02368 · Retrieved 19 November 2026Note. “We propose Visual Goal-conditioned Action Reasoning (ViGAR), a hierarchical framework that factorizes manipulation into a visual subgoal planner and a subgoal executor.” Hugging Face teaser describes “a subgoal planner and a world-action policy”.
- What it controls
- Arms and handsSource and note: What it controlsB dagroup-pku.github.io/ViGAR/ · Retrieved 19 November 2026Note. “Experiments run on an AgiBot A2 with two 7-DoF arms, two wrist cameras and one front-facing camera.”
- Inputs
- Camera LanguageSource and note: InputsA arxiv.org/abs/2610.02368 · Retrieved 19 November 2026Note. “Given the current observation and global instruction, the subgoal planner predicts a visual subgoal for the next subtask.”
- Access
- Open weights (download)Source and note: AccessC huggingface.co/DAGroup-PKU/ViGAR · Retrieved 19 November 2026Note. “Evaluation code and checkpoints on **RoboTwin**.”
- Base model
- Cosmos3-NanoSource and note: Base modelB dagroup-pku.github.io/ViGAR/ · Retrieved 19 November 2026Note. “Both share the Cosmos3-Nano reasoner–generator backbone.”
- Licence document
- OpenMDW-1.1Source and note: Licence documentC huggingface.co/DAGroup-PKU/ViGAR · Retrieved 19 November 2026Note. Hugging Face model card header: “license_name: openmdw-1.1”. License section: “Original project code uses the [MIT license](https://github.com/DAGroup-PKU/ViGAR/blob/main/LICENSE). Bundled components retain their upstream licenses. Cosmos-derived model materials use OpenMDW-1.1.”

Specifications
7 of 14 stated 7 not stated
5 fields hidden 2 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
Openness2/3 disclosed
The developer’s own claims0/2 disclosed
Notes
Developed by researchers from Peking University, AgiBot, and CocoMatrix.
Described by the authors as “We propose Visual Goal-conditioned Action Reasoning (ViGAR), a hierarchical framework that factorizes manipulation into a visual subgoal planner and a subgoal executor.”
Sources
One line per source, not per figure: a source has one retrieval date.
- Aarxiv.org/abs/2610.02368Retrieved 19 November 2026
- Bdagroup-pku.github.io/ViGAR/Main source. Retrieved 19 November 2026
- Chuggingface.co/DAGroup-PKU/ViGARRetrieved 19 November 2026
More models
Same developer

LAWPeking University
- No photograph
LDA-1BPeking University
- No photograph
MM-NavPeking University
- No photograph
TrackVLAPeking University
- No photograph
TrackVLA++Peking University

Uni-NaVidPeking University



