Subham Sahoo

@ssahoo_

Pioneering Diffusion LLMs | Team Lead @mbzuai - IFM | PhD @cornell

following
followers

Uno-adapters incorporated into Nvidia NeMo

A post celebrating the change thanks three people and links to more information about Uno.

2 Sources, 6d ago, first seen 6d ago

'AR vs Diffusion' panel planned for COLM on October 9

A prospective panelist said they looked forward to taking part in the COLM session that day.

3 Sources, 10d ago, updated 10d ago

Uno reportedly outperforms DFlash in @broad_net's use, especially at longer context lengths

A user says many inference service providers have expressed interest in adopting Uno to speed up LLM inference.

2 Sources, 19d ago, first seen 19d ago

Uno promises diffusion-style speed with autoregressive LLM quality

Uno’s developer says each layer has separate autoregressive and diffusion weights, enabling parallel sampling without changing the autoregressive output distribution.

9 Sources, 35d ago, first seen 35d ago

Subham Sahoo Posts 2B MDLM Models on HF

Subham Sahoo announces 2B MDLM, Duo and Eso-LMs models now on Hugging Face.

1 Source, 44d ago, first seen 44d ago

Paper Introduces Diffusion-Augmented LLMs

Tweet shares arXiv paper on diffusion-augmented LLMs for parallel token sampling.

17 Sources, 47d ago, first seen 47d ago

Featured in

Uno-adapters incorporated into Nvidia NeMo

'AR vs Diffusion' panel planned for COLM on October 9

Uno reportedly outperforms DFlash in @broad_net's use, especially at longer context lengths

Uno promises diffusion-style speed with autoregressive LLM quality

Subham Sahoo Posts 2B MDLM Models on HF

Paper Introduces Diffusion-Augmented LLMs

Featured in

Uno-adapters incorporated into Nvidia NeMo

'AR vs Diffusion' panel planned for COLM on October 9

Uno reportedly outperforms DFlash in @broad_net's use, especially at longer context lengths

Uno promises diffusion-style speed with autoregressive LLM quality

Subham Sahoo Posts 2B MDLM Models on HF

Paper Introduces Diffusion-Augmented LLMs