#FacebookResearch
We've open sourced Adjoint Sampling!

It's part of a bundled release showcasing FAIR's research and open source commitment to AI for science.

github.com/facebookrese...

There's also a blog about the release
ai.meta.com/blog/meta-fa...
GitHub - facebookresearch/adjoint_sampling: code for "Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matching"
code for "Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matching" - facebookresearch/adjoint_sampling
github.com
May 14, 2025 at 6:20 PM
Meta's DACVAE

VAE version of the Descript Audio Codec, which has a continuous latent space. Descript Audio Codec (DAC) is a high fidelity general neural audio codec.

github.com/facebookrese...
GitHub - facebookresearch/dacvae: DACVAE
DACVAE. Contribute to facebookresearch/dacvae development by creating an account on GitHub.
github.com
December 19, 2025 at 4:09 AM
Proud to share our work on Large Concept Models (LCMs)! This is a new direction in language modeling that moves beyond traditional token-level LLMs.
Paper: ai.meta.com/research/pub...
Code: github.com/facebookrese...
GitHub - facebookresearch/large_concept_model: Large Concept Models: Language modeling in a sentence representation space
Large Concept Models: Language modeling in a sentence representation space - facebookresearch/large_concept_model
github.com
December 14, 2024 at 6:59 PM
Meta just released a human-centric pretrained vision transformers (pose, segmentation, normals, and pointmaps).

github.com/facebookrese...

The repo was only opened 40 minutes ago, looks like it's so early the project page isn't even accessible.
GitHub - facebookresearch/sapiens2: [ICLR 26] 1K resolution vision transformers pretrained on 1B human images.
[ICLR 26] 1K resolution vision transformers pretrained on 1B human images. - facebookresearch/sapiens2
github.com
April 24, 2026 at 2:46 AM
Paper: SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution
( arxiv.org/abs/2502.18449 )
Repo: github.com/facebookrese...
GitHub - facebookresearch/swe-rl: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution
Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution - facebookresearch/swe-rl
github.com
February 26, 2025 at 6:40 AM
Meta's Lyapunov

PyTorch original implementation of "Global Lyapunov functions: a long-standing open problem in mathematics, with symbolic transformers"

Repo: github.com/facebookrese...
Paper: arxiv.org/abs/2410.08304
GitHub - facebookresearch/Lyapunov: PyTorch original implementation of "Lyapunov functions a long-standing open problem in mathematics, with symbolic transformers" (NeurIPS 2024).
PyTorch original implementation of "Lyapunov functions a long-standing open problem in mathematics, with symbolic transformers" (NeurIPS 2024). - facebookresearch/Lyapunov
github.com
December 7, 2024 at 6:57 AM
digit360

Digit 360 is a modular platform that unlocks new capabilities, and enables future research on the nature of touch.

https://github.com/facebookresearch/digit360
November 15, 2024 at 2:15 AM
Paper: R.I.P.: Better Models by Survival of the Fittest Prompts ( arxiv.org/abs/2501.18578 )

They used:
- Fairseq2 from FAIR for LLM finetuning ( github.com/facebookrese... )
- vLLM for inference ( github.com/vllm-project... )
GitHub - facebookresearch/fairseq2: FAIR Sequence Modeling Toolkit 2
FAIR Sequence Modeling Toolkit 2. Contribute to facebookresearch/fairseq2 development by creating an account on GitHub.
github.com
January 31, 2025 at 4:48 AM
Coconut

Training Large Language Model to Reason in a Continuous Latent Space.

https://github.com/facebookresearch/coconut
January 28, 2025 at 12:15 AM
And our package is MIT-licensed and it's on Github!
neuroai/neuralbench-repo at main · facebookresearch/neuroai
Python suite for neuroscience research across all modalities. - facebookresearch/neuroai
github.com
May 6, 2026 at 6:54 PM
flow_matching

A PyTorch library for implementing flow matching algorithms, featuring continuous and discrete flow matching implementations. It includes practical examples for both text and image modalities.

https://github.com/facebookresearch/flow_matching
December 18, 2024 at 9:15 PM
To confirm the origin of a video, the watermark and hidden message can be revealed to verify the source of the video.

Github: github.com/facebookrese...
GitHub - facebookresearch/videoseal: Open and efficient video watermarking
Open and efficient video watermarking. Contribute to facebookresearch/videoseal development by creating an account on GitHub.
github.com
December 14, 2024 at 3:07 PM
📦 facebookresearch / detectron2
⭐ 32,834 (+43)
🗒 Python

Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.
GitHub - facebookresearch/detectron2: Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.
Detectron2 is a platform for object detection, segmentation and other visual recognition tasks. - facebookresearch/detectron2
github.com
September 17, 2025 at 2:02 PM
In my PhD much of my code was hacky, and I think this set me back quite a bit. At some point I overcorrected towards building complex frameworks for my work, which let me try a lot of things (so long as I stayed within my own framework). This is more or less what you see in NeuralCompression.
GitHub - facebookresearch/NeuralCompression: A collection of tools for neural compression enthusiasts.
A collection of tools for neural compression enthusiasts. - facebookresearch/NeuralCompression
github.com
December 13, 2024 at 3:26 PM
It appears that modern deep learning optimizers Schedule Free, an alternative to momentum update that involves iterates averaging and SOAP, a Shampoo-style 2nd order preconditioned AdamW, combine well together:

github.com/facebookrese...
github.com/nikhilvyas/S...

x.com/_clashluke/s...
GitHub - facebookresearch/schedule_free: Schedule-Free Optimization in PyTorch
Schedule-Free Optimization in PyTorch. Contribute to facebookresearch/schedule_free development by creating an account on GitHub.
github.com
November 8, 2024 at 11:01 AM
Keep tabs on LCMs - they're showing exceptional promise for translation. The key is that the sentences (not words) are the unit-of-measure, the encoder is (say) "English to vector space", and the decoder is "vector space to Mandarin".

github.com/facebookrese...
GitHub - facebookresearch/large_concept_model: Large Concept Models: Language modeling in a sentence representation space
Large Concept Models: Language modeling in a sentence representation space - facebookresearch/large_concept_model
github.com
February 20, 2026 at 10:15 PM
MAE actually produces great embeddings for everything… in case you don’t mind the masking: github.com/facebookrese...
GitHub - facebookresearch/mae: PyTorch implementation of MAE https//arxiv.org/abs/2111.06377
PyTorch implementation of MAE https//arxiv.org/abs/2111.06377 - facebookresearch/mae
github.com
February 27, 2025 at 5:27 PM
Så blev Yann endelig glad. Ikke alt det token-fis. Vi vil have reasoning i embedding space! ✊🔥 Og al træningskoden er endda frit tilgængelig på GitHub 👇 Hvad tænker du om den her tilgang? 🤔👇 github.com/facebookrese... #dkai
GitHub - facebookresearch/large_concept_model: Large Concept Models: Language modeling in a sentence representation space
Large Concept Models: Language modeling in a sentence representation space - facebookresearch/large_concept_model
github.com
December 23, 2024 at 7:39 PM
7/ BLT has direct access to bytes but at its core still has a Transformer that operates over patches - the best of both worlds. But it’s just the beginning! We are releasing the training and inference code at github.com/facebookrese...
GitHub - facebookresearch/blt: Code for BLT research paper
Code for BLT research paper. Contribute to facebookresearch/blt development by creating an account on GitHub.
github.com
December 13, 2024 at 4:53 PM
There are efforts to remove this limitations, but so far tokenizers rule supreme.

See e.g. the "Byte Latent Transformer" -> github.com/facebookrese...
GitHub - facebookresearch/blt: Code for BLT research paper
Code for BLT research paper. Contribute to facebookresearch/blt development by creating an account on GitHub.
github.com
August 8, 2025 at 5:19 PM
OrienterNet
Visual Localization in 2D Public Maps
with Neural Matching #osint #geolocalization

https://github.com/facebookresearch/OrienterNet
July 3, 2023 at 9:09 AM