Newly published work from FAIR, Chameleon: Mixed-Modal Early-Fusion Foundation Models. This research presents a family of early-fusion token-based mixed-modal models capable of understanding & generating images & text in any arbitrary sequence. Paper ➡️ go.fb.me/7rb19n

Excited to release our work from last year showcasing a stable training recipe for fully token-based multi-modal early-fusion auto-regressive models! arxiv.org/abs/2405.09818 Huge shout out to Armen Aghajanyan Ramakanth Luke Zettlemoyer Gargi Ghosh and other co-authors. (1/n)

Super excited to introduce HippoRAG, a method I enjoyed developing the most in 2024. It’s led by my amazing student Bernal Bernal Jiménez and joint with Yiheng Shu Yu Gu Michi Yasunaga. Bernal’s thread gives a good technical account, so I’ll just share some personal thoughts

Curious about enhancing factuality and attribution in LLM generation? Check out our paper: arxiv.org/abs/2405.19325 Introducing NEST🪺: Nearest Neighbor Speculative Decoding for LLM Generation and Attribution, a training-free method that adds real-world texts into LLM generation.

Open sourcing Chameleon! Our work from last year - early fusion multimodal foundation model. We are releasing multimodalality in the input with text generation in the output( though the model was trained to generate text and image).

Last week we released Meta Chameleon: a new mixed-modal research model from Meta FAIR. Get the models ➡️ go.fb.me/4m87kk The 7B & 34B safety tuned models we’ve released can take any combination of text and images as input and produce text outputs using a new early

🚀💡We're hiring interns for 2025 at FAIR @ AI at Meta Work on cutting-edge projects: social reasoning, alignment, interaction, multi-agent communication & more with text/multimodal LLMs. Apply now! 🔗metacareers.com/jobs/119904986…

New from Meta FAIR — 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 ➡️ go.fb.me/w23lmz

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!

Last one of the year - EWE: arxiv.org/pdf/2412.18069 Ewe (Explicit Working Memory), enhances factuality in long-form text generation by integrating a working memory that receives real-time feedback from external resources.

New research from Meta FAIR — Meta Memory Layers at Scale. This work takes memory layers beyond proof-of-concept, proving their utility at contemporary scale ➡️ go.fb.me/3lbt4m

Meet ReasonIR-8B✨the first retriever specifically trained for reasoning tasks! Our challenging synthetic training data unlocks SOTA scores on reasoning IR and RAG benchmarks. ReasonIR-8B ranks 1st on BRIGHT and outperforms search engine and retriever baselines on MMLU and GPQA🔥

Congratulations to Sewon Min for winning yet another dissertation award, this time, ACL dissertation award. So proud of you Sewon!!!

🌿Introducing MetaCLIP 2 🌿 📝: arxiv.org/abs/2507.22062 code, model: github.com/facebookresear… After four years of advancements in English-centric CLIP development, MetaCLIP 2 is now taking the next step: scaling CLIP to worldwide data. The effort addresses long-standing

...is today a good day for new paper posts? 🤖Learning to Reason for Factuality 🤖 📝: arxiv.org/abs/2508.05618 - New reward func for GRPO training of long CoTs for *factuality* - Design stops reward hacking by favoring precision, detail AND quality - Improves base model across

🔍 How do we teach an LLM to 𝘮𝘢𝘴𝘵𝘦𝘳 a body of knowledge? In new work with AI at Meta, we propose Active Reading 📙: a way for models to teach themselves new things by self-studying their training data. Results: * 𝟔𝟔% on SimpleQA w/ an 8B model by studying the wikipedia

New research from FAIR- Active Reading: a framework to learn a given set of material with self-generated learning strategies for generalized and expert domains(such as Finance). Absorb significantly more knowledge than vanilla finetuning and usual data augmentations strategies

🤔Ever wondered why your post-training methods (SFT/RL) make LLMs reluctant to say “I don't know?” 🤩Introducing TruthRL — a truthfulness-driven RL method that significantly reduces hallucinations while achieving accuracy and proper abstention! 📃arxiv.org/abs/2509.25760 🧵[1/n]

New paper 📢 Most powerful vision-language (VL) reasoning datasets remain proprietary 🔒, hindering efforts to study their principles and develop similarly effective datasets in the open 🔓. Thus, we introduce HoneyBee, a 2.5M-example dataset created through careful data