Xinlei Chen

Researcher@Meta·美国·OpenReview

研究方向

self-supervised learning · unsupervised learning · visual question answering · instance segmentation · object detection · embodied agent · context reasoning

Xinlei Chen

Meta CLIP 2: A Worldwide Scaling Recipe

Learning to (Learn at Test Time): RNNs with Expressive Hidden States

Deconstructing Denoising Diffusion Models for Self-Supervised Learning

An Image is Worth More Than 16x16 Patches: Exploring Transformers on Individual Pixels

Learning to (Learn at Test Time): RNNs with Expressive Hidden States

Highly Compressed Tokenizer Can Generate Without Training

Learnings from Scaling Visual Tokenizers for Reconstruction and Generation

Gaussian Masked Autoencoders

LLMs can see and hear without any training

Deep Neural Networks without Normalization

Video Representation Learning Without Natural Videos

Massive Activations in Large Language Models

R-MAE: Regions Meet Masked Autoencoders

Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers

On the Surprising Effectiveness of Attention Transfer for Vision Transformers

V-JEPA: Latent Video Prediction for Visual Representation Learning

Learning to (Learn at Test Time)