Benjamin Muller
@ben_mlr
Followers
968
Following
2K
Media
14
Statuses
190
Research in AI. Focusing on scaling language models multi-modally & multilingually @AIatMeta
NYC
Joined April 2016
So many exciting releases from FAIR @AIatMeta Super happy to see Spirit LM now open-sourced. Spirit LM unlocks expressive speech generation through interleaving speech-text training and phonetic(hubert)+pitch+style-specific tokenization. Available here: Weights:
Open science is how we continue to push technology forward and today at Meta FAIR weโre sharing eight new AI research artifacts including new models, datasets and code to inspire innovation in the community. More in the video from @jpineau1. This work is another important step
1
1
14
๐ Introducing the Latent Speech-Text Transformer (LST) โ a speech-text model that organizes speech tokens into latent patches for better textโspeech transfer, enabling steeper scaling laws and more efficient multimodal training โก๏ธ Paper ๐ https://t.co/4nUsbC1YKF
7
17
34
Introducing @CodeWordsAI , the fastest way to go from idea to automation, simply by chatting with AI. No more drag-and-drop and configuration. Save time by doing less. Available today for free, for everyone. The Cursor moment for automation is here.
7
14
47
Thrilled to share that our Byte Latent Transformer won an Outstanding Paper Award at ACL 2025! ๐
๐ Introducing the Byte Latent Transformer (BLT) โ An LLM architecture that scales better than Llama 3 using byte-patches instead of tokens ๐คฏ Paper ๐ https://t.co/5QGrlJdK0y Code ๐ ๏ธ https://t.co/jCdDI5BXwe
16
31
282
We ran Llama 4 Maverick through some HELM benchmarks. It is 1st on HELM capabilities (MMLU-Pro, GPQA, IFEval, WildBench, Omni-MATH), butโฆ https://t.co/uKMHRe7xKF
5
18
140
Today is the start of a new era of natively multimodal AI innovation. Today, weโre introducing the first Llama 4 models: Llama 4 Scout and Llama 4 Maverick โ our most advanced models yet and the best in their class for multimodality. Llama 4 Scout โขย 17B-active-parameter model
835
2K
13K
๐จ Diverse Preference Optimization (DivPO) ๐จ SOTA LLMs have model collapse๐ซ : they can't generate diverse creative writing or synthetic data ๐จ DivPO trains for both high reward & diversity, vastly improving variety with similar quality. Paper ๐: https://t.co/bRwq3d3wJq ๐งตbelow
1
78
343
We released new research - Byte Latent Transformer(BLT) BLT encodes bytes into dynamic patches using light-weight local models and processes them with a large latent transformer. Think of it as a transformer sandwich!
New from Meta FAIR โ Byte Latent Transformer: Patches Scale Better Than Tokens introduces BLT, which for the first time, matches tokenization-based LLM performance at scale with significant improvements in inference efficiency & robustness. Paper โก๏ธ https://t.co/0iamZCRnMN
11
85
661
New from Meta FAIR โ Byte Latent Transformer: Patches Scale Better Than Tokens introduces BLT, which for the first time, matches tokenization-based LLM performance at scale with significant improvements in inference efficiency & robustness. Paper โก๏ธ https://t.co/0iamZCRnMN
28
192
1K
Groundbreaking scaling trends for Byte-level Language Modeling with the new BLT architecture ๐ More insights in the thread ๐งต
๐ Introducing the Byte Latent Transformer (BLT) โ An LLM architecture that scales better than Llama 3 using byte-patches instead of tokens ๐คฏ Paper ๐ https://t.co/5QGrlJdK0y Code ๐ ๏ธ https://t.co/jCdDI5BXwe
0
4
21
Congrats @aymericzzz and team on being live! Very exciting vision to build entire softwares with just a prompt
Excited to share more about our background, vision and where we're headed at @agemoai with @r1ddhi at @BusinessInsider ๐ข๐๐ฟ ๐๐ถ๐๐ถ๐ผ๐ป ๐ถ๐ ๐๐ผ ๐ฒ๐ป๐ฎ๐ฏ๐น๐ฒ ๐ฎ๐ป๐๐ผ๐ป๐ฒ ๐๐ผ ๐ฐ๐ฟ๐ฒ๐ฎ๐๐ฒ ๐๐ผ๐ณ๐๐๐ฎ๐ฟ๐ฒ โ from an idea to fully deployed software. The critical path to
0
2
4
๐ Iโve always had a dream of making AI accessible to everyone, regardless of location or language. However, current open MLLMs often respond in English, even to non-English queries! ๐ Introducing Pangea: A Fully Open Multilingual Multimodal LLM supporting 39 languages! ๐โจ
7
81
378
Meta Spirit LM: open source language model that mixes text and speech.
Today we released Meta Spirit LM โ our first open source multimodal language model that freely mixes text and speech. Many existing AI voice experiences today use ASR to techniques to process speech before synthesizing with an LLM to generate text โ but these approaches
19
72
331
Recent LLMs (e.g. LLama 3 ๐ฆ) are increasingly good at Math. However, this progress is reserved for languages with large amounts of task-specific instruct-tuning data. In this work @AIatMeta (led by @LucasBandarkar ), we introduce a new model merging technique called **Layer
Cross-lingual transfer can be as easy as swapping model layers between LLMs! ๐ Our model merging method can compose math and language skills by swapping top&bottom layers from a SFTโd target language expert into a math expert without retraining https://t.co/IN5JPdTYU4 ๐งต: [1/3]
2
7
28
OK here goes the "excited to share ...." post Want to know how to train a T2V model (with other amazing capabilities) that beats ALL prior work ?? Well we released a 90 page tech report with every detail ๐ https://t.co/FU2PzloDhrโฆ Thanks to the amazing team!
ai.meta.com
Meta Movie Gen is our latest research breakthrough that allows you to use simple text inputs to create videos and sounds, edit existing videos or transform your personal image into a unique video.
๐ฅ Today weโre premiering Meta Movie Gen: the most advanced media foundation models to-date. Developed by AI research teams at Meta, Movie Gen delivers state-of-the-art results across a range of capabilities. Weโre excited for the potential of this line of research to usher in
10
17
178
Introducing *Transfusion* - a unified approach for training models that can generate both text and images. https://t.co/h9PyPl1zNc Transfusion combines language modeling (next token prediction) with diffusion to train a single transformer over mixed-modality sequences. This
24
215
1K
LLM Evaluations are an important area of work โ today we're announcing a new LLM Evaluation Research Grant to foster further innovation in this area. Recipients will get $200K in funding to support this work. We're accepting proposals until September 6 โก๏ธ https://t.co/0tJcAFq4RO
15
90
456
Starting today, open source is leading the way. Introducing Llama 3.1: Our most capable models yet. Today weโre releasing a collection of new Llama 3.1 models including our long awaited 405B. These models deliver improved reasoning capabilities, a larger 128K token context
264
1K
6K