Back to feed
News Story
APriority78
Meta AI (FAIR)
1 sources

Learn More

Original

This is a generic placeholder with no real news content.

SynthePulse Insight · AI deep reading

TRIBE v2: A Digital Twin of Brain Activity, a Predictive Foundation Model for Neuroscience

Version 1 · 1 source

Meta FAIR releases TRIBE v2, a high-resolution model that predicts human brain responses to visual, auditory, and linguistic stimuli, with 70x resolution improvement and zero-shot prediction, aiming to accelerate neuroscience research and clinical breakthroughs.

  • TRIBE v2 is Meta FAIR's next-generation predictive foundation model, serving as a digital twin of human neural activity, predicting brain responses to almost any audiovisual stimulus with 70x higher resolution.
  • The model is trained on fMRI data from over 700 healthy volunteers exposed to various media including images, podcasts, videos, and text.
  • TRIBE v2 supports zero-shot prediction for new subjects, new languages, and new tasks, consistently outperforming standard modeling approaches.
  • The model, code, paper, and interactive demo are released under a CC BY-NC license for researchers.
  • The model aims to help neuroscientists and clinical researchers test theories without human subjects, accelerating breakthroughs in neurological disease treatments.
Open section navigationCore Breakthrough: From Low-Resolution Individual to High-Resolution Population Model

Core Breakthrough: From Low-Resolution Individual to High-Resolution Population Model

TRIBE v2 builds on Meta's previous award-winning model from Algonauts 2025. According to the Meta AI official blog, that award-winning model was trained on low-resolution fMRI recordings from only four individuals. In contrast, TRIBE v2 leverages a large-scale dataset of over 700 healthy volunteers presented with various media, including images, podcasts, videos, and text.

Meta claims that TRIBE v2 reliably predicts high-resolution fMRI brain activity with a 70-fold resolution improvement over similar models. This enhancement enables the model to achieve zero-shot prediction for new subjects, new languages, and new tasks, consistently outperforming standard modeling methods.

Application Prospects: Accelerating Neuroscience Discovery and Clinical Practice

Meta states that TRIBE v2, as a digital twin of human neural activity, allows neuroscientists and clinical researchers to rapidly test hypotheses about brain function without using human subjects. This is expected to accelerate understanding and treatment breakthroughs for neurological diseases affecting hundreds of millions of people.

Additionally, Meta notes that by applying brain insights to build better AI systems, the model can directly guide AI development with neuroscientific principles.

Openness and Accessibility

Meta has released the research paper, model weights, and code under a CC BY-NC license, along with an interactive demo website. Researchers can freely explore TRIBE v2's capabilities, thereby advancing cutting-edge neuroscience research.

Credibility boundary

All information in this article comes from the Meta AI official blog, a first-party source. All claims about model performance, dataset size, and resolution improvement are Meta's own assertions and have not been independently verified by third parties.

Insight takeaway

TRIBE v2 represents a significant advance in neuroscience AI from Meta, providing an unprecedented tool for understanding how the brain processes complex stimuli through large-scale data and zero-shot prediction capabilities. However, its actual effectiveness and clinical value still require independent validation.

Sources for this version

  1. Learn MoreMeta AI (FAIR)

Primary report

Meta AI (FAIR)

Primary source