Meta Research

Self-referential self-improving agents that can optimize for any computable task

This repository contains the code to train and evaluate TRIBE v2, a multimodal model for brain response prediction

Code for the ShapeR research paper

Efficient Universal Perception Encoder: a single on-device vision encoder with versatile representations that match or exceed specialized experts across multiple task domains.

A Large-scale Video Action Dataset

🎬ActionMesh: A fast video to animated mesh model with unprecedented quality. Generate animated mesh seamlessly importable into any 3D software in less than a minute.

Code for the Boxer research paper

A Python framework for AI-driven character animation using neural networks.

Official code for "LagerNVS Latent Geometry for Fully Neural Real-time Novel View Synthesis" (CVPR 2026)

A pure-Python implementation of the Nvidia CuTe layout algebra intended to be approachable and easy to learn.

Research code base for Automatic Textbook Formalization

AIRS-Bench: an AI Research Science benchmark for quantifying the end-to-end AI research abilities of LLM agents

Official code and data from DexWM ("World Models Can Leverage Human Videos for Dexterous Manipulation").

PyTorch Implementation of Image Generation with a Sphere Encoder

Code for "Agentic Very Long Video Understanding" (EGAgent) [ACL 2026 Main]

Code for Latent Speech-Text Transformer (LST)

Automatic textbook formalization of Grinberg Algebraic Combinatorics

Procedural Knowledge at Scale Improves ReasoningThis repository contains the minimal, end-to-end pipeline for reproducing the paper results generate a procedural knowledge datastore, build retrieval indices, run retrieval, perform model rollouts with retrieved subroutines, and filter the samples to output the final metrics.

Authors implementation of "Flowception Temporally Expansive Flow Matching for Video Generation".

A tutorial and a set of tools to compute depth-from-stereo with Project Aria Gen2 devices. This includes stereo image rectification as well as disparity estimation

WybeCoder Verified Generation of Imperative Code with LLMs

Bayesian Inference & Tooling for Experimentation Systems

MRS.Q is a model-based reinforcement learning algorithm that selects actions with search.

Physics-based bone-driven neural garment animation

Code for VoxelCodeBench.

MuLoCo Muon is a practical inner optimizer for DiLoCo

High-performance GPU kernels for Ads and Recsys model training, independently implemented and optimized for real-world workloads and model-specific input characteristics.

GISTBench Dataset and Evaluation Code

Implementation of the proposed algorithms and simulation studies in Adaptive Exploration for Latent-State Bandits

We propose a new spatio-directional neural encoding for Neural Rendering applications that is compact and efficient, and supports all-frequency signals in both space and direction.