Meta FAIR Unveils New AI Research and Open-Source Releases

Sharing new research, models, and datasets from Meta FAIR

We aim to democratize access to state-of-the-art technologies that transform our interaction with the physical world, which is why we're committed to fostering a collaborative and open ecosystem that accelerates progress and discovery. As we continue to work towards our goal of achieving advanced machine intelligence, we want to share our progress with the research community so they can build upon our work. Today, we're excited to release some of the latest research, code, models, and datasets from Meta Fundamental AI Research (FAIR). The artifacts we're sharing today focus on building more capable agents, robustness and safety, and architecture innovations that enable models to learn new information more effectively and scale beyond current limits. In this release, we're sharing a demo and code for Meta Video Seal, an open source model work video watermarking that builds on the popular Meta Audio Seal work we shared last year. We're also sharing a variety of other artifacts, including a foundation model for controlling the behavior of virtual embodied agents, a method for scaling memory layers that will enable more factual information, and code to help models become more socially intelligent. There's plenty more to explore in this post with nine total projectsand artifacts ready for people to download and start using today. This work supports our long and proven track record of sharing open reproducible science with the community. By publicly sharing our early research work, we hope to inspire iterations and ultimately help advance AI in a responsible way. As always, we look forward to seeing what the community will build using these new releases and continuing the dialogue about how we can all advance AI together responsibly and build for the greater good.

Meta FAIR Announces New Research Artifacts | AI News

New research showcase innovations in agents and machine learning architectures. Meta FAIR (The Fundamental AI Research) has unveiled new research artifacts, amidst major AI announcements from OpenAI and Google, according to their blog. It includes releases in language, embodied AI, new architectures, and more. "It's been a big year for AI, and today at NeurIPS I'm excited to share nine new open source releases from Meta FAIR to wrap up the year -- all part of our continued mission to achieve advanced machine intelligence (AMI), and I am very excited for even more to come in 2025! " said Joelle Pineau, VP of AI research, Meta, on Linkedin. Meta Motivo enables virtual humanoid agents to perform complex tasks with human-like behaviours by learning from motion datasets and adapting to environmental factors. Theory-of-Mind Data provides tools and datasets designed to train AI in understanding and predicting human thoughts and beliefs, significantly advancing research in social intelligence. Memory Layers and Large Concept Models (LCM) further enhance AI's capabilities by improving its ability to store, retrieve, and process complex information, enabling better handling of diverse languages and hierarchical reasoning. Complementing these advancements, Meta CLIP 1.2 is a vision-language model that precisely aligns image and text data, supporting critical tasks such as retrieval, classification, and multi-modal embedding. Meta Video Seal enhances digital content security by embedding robust, invisible watermarks in videos, ensuring they remain resistant to editing and compression. Supporting this innovation, Omni Seal Bench introduces a leaderboard for evaluating neural watermarking techniques, with plans for a dedicated workshop in 2025 to further advance the field. Additionally, Flow Matching serves as an open-source framework that facilitates the creation of high-quality images, videos, audio, and 3D structures, empowering users to generate rich and dynamic media content efficiently. On the whole, under Meta AI chief Yann LeCun's leadership, the company is advancing AI with systems like Layer Skip and V-JEPA to improve reasoning and interaction, alongside self-supervised learning to ultimately AMI.