Sucheng Ren

Hi, I am Sucheng Ren (任苏成), a Computer Science Ph.D. student at Johns Hopkins University, where I am fortunate to be advised by Professor Alan Yuille and Prof. Cihang Xie. I received my B.S. and M.S. degree in Computer Science from South China University of Technology advised by Prof. Shengfeng He. Currently, I am a research intern at Apple. Previously, I spent great time at Bytedance Seed, Microsoft Research Asia (MSRA), Tsinghua University and National University of Singapore.

My research lies at the Diffusion/Autoregressive based Generative Model and Multimodal Learning.

Email  |  CV  |  Scholar  |  Github  | 

We decouple scale-wise attention which allows to rebuild VAR in a more computationally efficient manner.

We explicitly incorporates frequency-aware conditioning into the flow matching.

We generalize next token prediction to next X prediction.

We generalize next scale prediction to simplest scale design and make it compatible with any VAE.

We are the first to pretrain Mamba in vision with Cluster-based autoregressive modeling

We use autoregressive pretraining for self-supervised video representation learning

We enhance image-GPT with semantic-rich supervision.

We explore distillation techniques to transfer the success of large MIM-based pre-trained models to smaller ones.

Integrating the capability of capturing multiscale objects in each attention layer by adaptively merging tokens.

Integrating the capability of capturing multiscale objects in each attention layer by adaptively merging tokens.

The first work delves into the influence of models inductive biases in knowledge distillation

A generic training strategy in data mixing that can improve the self-supervised representation learning of both CNNs and ViTs

Training a master to learn how to teach a better student.

Jointly learning salient objects, moving objects, recurring objects for Unsupervised Video Object Segmentation.