We previously shared our research on Layer Skip, an end-to-end solution for accelerating LLMs from researchers at Meta FAIR. It achieves this by executing a subset of an LLM’s layers and utilizing subsequent layers for verification and correction. We’re now releasing inference

Our brief overview of the audio generation model behind NotebookLM Audio Overviews and Illuminate.

Continuous Speech Synthesis using per-token Latent Diffusion Author's Explanation: x.com/AvihuDkl/statu… Overview: SALAD is introduced as a per-token latent diffusion model for zero-shot text-to-speech synthesis, focusing on continuous representations to improve

משלוחים מאלי אקספרס: שומע אחי פחות זרם לנו להגיע לרמת גן אז שמנו לך את זה בbox בקריית שמונה תבוא לאסוף תוך 48 שעות שים גז השעון כבר מתקתק.

Ever wanted to locate your cat in a database of images using just one reference image? Probably not—but this highlights a gap in VLMs. They struggle to localize specific objects given in-context examples, often copying the last sample's location instead of learning from it.

Kudos to the fantastic team! Impressive results!

Say I want to compare system qualities - pick between 2 configurations, or rank a whole bunch of models. I'll use LLM-as-a-judge, right? 🧑🏻‍⚖️ But how do I know the LLM judge is up to the task? Who is a good judge for ranking systems? Enter our new paper!✨🧵 arxiv.org/abs/2412.09569

המשחק היומי: רעמים או בומים

רעמים וחות׳ים בליל חורף קר לא נשמעים תמיד אותו הדבר

[1/5] Rethinking SVGs, the implicit way — meet NeuralSVG! 🎨✨ An implicit neural representation for generating layered SVGs from text prompts. 💧 Powered by SDS and nested-dropout for ordered shapes 🖌️ Enables inference-time editing like color palette & aspect ratio Read more on

IBM released Granite-Vision-3.1-2B, a small vision LM with impressive performance on different tasks 😮🔥 it comes with transformers and vLLM support from the get-go 💗 you can run it in Colab T4, so I built a notebook to put it to test, find it on the next one ⤵️

📣 We've been cooking something special... I'm excited to share #GraniteVision from IBM Research - a compact 2B parameter vision-language model that's "punching above its weight(s)" in visual document understanding! Small model, smart data, big results! 💪 #AI #VLM #Multimodal

Love the figures 😍 👏

🚀 Happy to share that we released Granite Speech 3.3 8B. It's a transcription and translation model that outperforms leading ASR systems. Check it out, it's open-source (Apache 2.0). ibm.com/new/announceme…

הרגע ראיתי את הנהג מדלג על תחנה שיש בה נוסעים שמנופפים לו מה עושים?

חסרים לנו כמה למניין

We released a 2B version of Granite Speech, for cost-efficient inference. Check it out: huggingface.co/ibm-granite/gr…

We are in the era of AI. How can I read aloud a PDF? Microsoft edge was pretty good, but it no longer works 😕

Feels good to be at the top of a leaderboard :)

בוקר יום ראשון טוב לכולם