Downloads · 30 days
17
13% of all-time downloads
ajaysri/prism-qwen25-extra-dinosiglip-224px-0_5b-stage-finetune-x7
prism-qwen25-extra-dinosiglip-224px-0_5b-stage-finetune-x7 is a robotics model from ajaysri. Use it for the robotics task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Base vision-language model used to initialize the public BARX policy checkpoints.
Downloads · 30 days
17
13% of all-time downloads
All-time downloads
131
Public
Repo size
7.9 GB
Likes
0
Public
Click a slice to open those files.
.pt10.5 GB · 99%
From the Hugging Face model README
Base vision-language model used to initialize the public BARX policy checkpoints.
This repository supplies config.json and
checkpoints/latest-checkpoint.pt for the BARX release manifest. It is a base
VLM, not a directly deployable robot policy; task-specific VQ tokenization,
normalization, training, and evaluation code live in the
BARX release repository.
The repository does not independently restate or certify the upstream pretraining corpora of Qwen2.5, DINOv2, or SigLIP. Consult their model cards for upstream data and limitations.
Use this artifact to reproduce or extend the BARX simulation-policy training pipeline. It is not validated for direct real-robot control, safety-critical deployment, or generic image/text generation. Public release smoke tests establish loadability and trainability, not paper-scale performance.
This BARX-composed model artifact is released under Apache 2.0. Preserve the attribution and separately applicable terms for Qwen2.5, DINOv2, SigLIP, MiniVLA/OpenVLA, RoboCasa/RoboCasa-X, and BARX.
@inproceedings{sridhar2026barx,
title = {Cross-Embodiment Transfer via Behavior-Aligned Representations},
author = {Sridhar, Ajay and Gao, Jensen and Yang, Jonathan and Mercat, Jean and Belkhale, Suneel and Sadigh, Dorsa},
booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
year = {2026}
}