Our own vasudevlal highlights RAG's role in advancing AI with less effort. RAG enables smarter, data-driven AI outputs without the need for extensive training.

Explore his insights on its strategic application: linkedin.com/pulse/when-use…

Getting that cat 'next' to a suitcase seems to be always a problem for T2I diffusion models 🤷‍♂️

In our latest work, we investigate the problem of spatial consistency, improve upon it, and offer multiple insights into solving it with systematic re-captioning methods.

We make our

LDM3D-VR: Latent Diffusion Model for 3D VR

abs: arxiv.org/abs/2311.03226

This paper from Intel Labs introduces a suite of diffusion models for VR development, include a panoramic generation model finetuned from Stable Diffusion v1.5 and a super-resolution model finetuned from

Extremely proud that our paper (LDM3D-VR) Latent Diffusion Model for 3D VR was featured by AK in Hugging Face's 'Daily Papers' selection for Nov 7.

Check out the paper!

Hugging Face Daily papers email of 7 Nov 2023 is out

huggingface.co/papers

Intel based accelerators deliver 2.5x permanece over $NVDA A100 and 1.4x over the H1 on Hugging Face vision language model.

ASIC versus GPU not 1:1 but an encouraging data pout for @Intel as it ramps up its #AI efforts.

huggingface.co/blog/bridgetow…

$INTC

We just released a new benchmark where Habana Labs Gaudi2 is up to x1.4 faster than Nvidia H100!
x1.2 speedup with the exact same training arguments, and x1.4 with hardware-accelerated data loading which can be done easily with Optimum Habana.
More here: huggingface.co/blog/bridgetow…

New ML gem on the hub: LDM3D by Intel. This diffusion model generates image & depth from text prompts. Using a custom @gradio 6dof three.js component you can generate immersive 360-degree views from prompts

demo: huggingface.co/spaces/Intel/l…
model: huggingface.co/Intel/ldm3d-pa…

How can we build better counterfactuals for data augmentation?

Check out our #emnlp2022 paper: 'NeuroCounterfactuals: Beyond Minimal-Edit Counterfactuals for Richer Data Augmentation'(arxiv.org/abs/2210.12365) w/ Phillip Howard, Gadi Singer, vasudevlal and Yejin Choi
🧵1/

Have you ever wanted to create your own world?

Intel and Blockade Labs are using #GenerativeAI to make immersive 360-degree views from simple text prompts.

Learn more: intel.ly/3pcUzU9

In collaboration with @Intel, our Depth Fusion showcases the power of our LDM3D diffusion model in generating 360° views from text prompts provided by the user.

The LDM3D diffusion model generates a 2D RGB image and its corresponding relative depth map providing a complete RGBD

congrats to Amir David Nissan cohen and Shachar Rosenman for winning the best paper award at AKBC 2022! (for a paper that was previously rejected from multiple NLP venues...)

very excited to share our paper on reconstructing language from non-invasive brain recordings! we introduce a decoder that takes in fMRI recordings and generates continuous language descriptions of perceived speech, imagined speech, and possibly much more biorxiv.org/content/10.110…

VL-InterpreT: An Interactive Visualization Tool for Interpreting Vision-Language Transformers
abs: arxiv.org/abs/2203.17247

Best Demo Award #2:
VL-InterpreT: An Interactive Visualization Tool for Interpreting Vision-Language Transformers by Meng Du @shaoyenT vasudevlal from Intel Labs, Intel AI & UCLA
Code & Demo: github.com/IntelLabs/VL-I…

I'll be presenting our VL-InterpreT demo #CVPR2024 ! We developed an interactive tool to visualize the attention, attention head statistics, and hidden representations in vision-language transformers. You can use this tool with any vision-language transformer. #CVPR2022

From attention to consciousness: Michael Graziano, Jonathan Cohen, Vasudev Lal and me will discuss about attention models in philosophy, cognitive science, neuroscience and AI, this Thursday Jan 20, 9am EST; free tickets: eventbrite.com/e/vectors-of-c…

#nlproc we have a stellar list of speakers for #kinn2021 #cikm2021 workshop: Luis Lamb Subbarao Kambhampati (కంభంపాటి సుబ్బారావు) Gary Marcus Vered Shwartz Antoine Bosselut Partha Talukdar

Regular papers/non-archival submissions/WIP welcome.
sites.google.com/view/kinn2021/…

Pasquale Minervini 🚀 looking for postdocs! Somak Aditya @Yezhou_Yang vasudevlal