EfficientSAM, small but mighty! With 20x fewer parameters and 20x faster runtime, EfficientSAM is within 2 points of the original SAM model, outperforming MobileSAM/FastSAM by a large margin. Paper: arxiv.org/pdf/2312.00863… Project details and demo: yformer.github.io/efficient-sam/

MobileLLM, our recent on-device efforts for large language models (LLMs with 125M/350M parameters)

The guide of vision-language models (VLMs).

MobileLLM: nice paper from AI at Meta about running sub-billion LLMs on smartphones and other edge devices. TL;DR: more depth, not width; shared matrices for token->embedding and embedding->token; shared weights between multiple transformer blocks; Paper: arxiv.org/abs/2402.14905

Agent-as-a Judge: our new work on using AI agents to evaluate AI agents, achieving human-level performance.

🚨VideoLLM from Meta!🚨 LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding 📝Paper: huggingface.co/papers/2410.17… 🧑🏻‍💻Code: github.com/Vision-CAIR/Lo… 🚀Project (Demo): vision-cair.github.io/LongVU We propose LongVU, a video LLM with a spatiotemporal adaptive

🚀We're thrilled to announce the MobileLLM weights are Available on HuggingFace: huggingface.co/collections/fa… 📱MobileLLM is a state-of-the-art language model designed for mobile devices： arxiv.org/abs/2402.14905 🔥Explore the pretraining code on GitHub: github.com/facebookresear…

[1/n] 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗧𝗿𝗮𝗰𝗸 𝗔𝗻𝘆𝘁𝗵𝗶𝗻𝗴 from Meta: interactive video segmentation and tracking on an iPhone!

Efficient Track Anything for segment everything 🔥 Gradio demo (built upon SkalskiP 's sam2 demo): 5239f8e221db7ee8a0.gradio.live #supermario #UniversalStudios #sanfranisco

Efficient Track Anything is on alphaxiv.org/abs/2411.18933. I am happy to take questions on alpharxiv. alphaXiv

Mitigating racial bias from LLMs is a lot easier than removing it from humans! Can’t believe this happened at the best AI conference NeurIPS Conference We have ethical reviews for authors, but missed it for invited speakers? 😡

This project really changed how I think about multimodal models and LLMs. I used to believe that multimodal (visual) prediction required significant changes to the model and heavy pretraining, like Chameleon. But surprisingly, the opposite is true! In large autoregressive models,

🚀MetaMorph can leverage LLM knowledge for generation and implicit reasoning.

Excited to see Peter Tong 's internship work (MetaMorph) on exploring unified multimodal understanding and generation with many interesting findings. Check out the paper and project page below.

Glad to see efficient track anything model has been used for near real-time multi-object segment and tracking on a MacBook for SlapFX. Looking forward to your next-gen CapCut release, Hart Woolery !

ParetoQ, unified LLM quantization framework across all bit-width levels.

Efficient Track Anything is accepted to ICCV 2025! See you in Hawaii!