Back to Organizations
Bagel Labs
Bagel Labs is a physical AI research lab developing compact world-action models and distributed training systems for autonomous robot control.
San Francisco, California, United States; Toronto, Ontario, Canada
The Bagel Network
Project Links
Description
Bagel Network is a decentralized data platform dedicated to supporting machine learning (ML) models by addressing the issue of concentrated data ownership in the ML domain. It aims to establish a marketplace where data scientists and AI engineers can efficiently exchange and license verifiable datasets in a privacy-preserving manner, fostering collaboration and accessibility in the machine learning community. The platform operates as a standalone chain, utilizing zero-knowledge commitments, reputation mechanisms, and staked tokens to ensure data integrity and quality.Category: AI
PARIS 1.0
Project Links
Description
PARIS is Bagel Labs’ training architecture for physical AI and robotics. It enables different parts of a robotics model to learn independently and then combines them into one model. The Bagel Labs site reports a 24% improvement for PARIS 1.0 in matched-resource image- and video-generation tests, while explicitly noting that these results are not robotics validation. PARIS is presented separately from WorldDiT, Bagel’s model architecture for robot learning and control.Category: Uncategorized
WorldDiT
Project Links
Description
WorldDiT jointly generates continuous robot-action chunks and predicts future normalized RGB patches through one diffusion transformer. Bagel Labs publicly released four LIBERO checkpoints, a self-contained inference runtime, and an evaluator under a CC BY 4.0 license.Category: Uncategorized
Paris
Project Links
Description
Paris is a decentralized-trained diffusion model comprising independently trained expert models and an inference router. The project was released as an open-weight text-to-image model under the MIT license; Bagel Labs later announced Paris 2.0 for video generation using the same decentralized-training approach.Category: Uncategorized