Výsledky vyhledávání - "Kong Xiangwen"

Report

DreamLLM: Synergistic Multimodal Comprehension and Creation

Autor: Dong, Runpei, Han, Chunrui, Peng, Yuang, Qi, Zekun, Ge, Zheng, Yang, Jinrong, Zhao, Liang, Sun, Jianjian, Zhou, Hongyu, Wei, Haoran, Kong, Xiangwen, Zhang, Xiangyu, Ma, Kaisheng, Yi, Li

This paper presents DreamLLM, a learning framework that first achieves versatile Multimodal Large Language Models (MLLMs) empowered with frequently overlooked synergy between multimodal comprehension and creation. DreamLLM operates on two fundamental

Externí odkaz: http://arxiv.org/abs/2309.11499

Zobrazit plný text záznamu

Report

Reversible Column Networks

Autor: Cai, Yuxuan, Zhou, Yizhuang, Han, Qi, Sun, Jianjian, Kong, Xiangwen, Li, Jun, Zhang, Xiangyu

We propose a new neural network design paradigm Reversible Column Network (RevCol). The main body of RevCol is composed of multiple copies of subnetworks, named columns respectively, between which multi-level reversible connections are employed. Such

Externí odkaz: http://arxiv.org/abs/2212.11696

Zobrazit plný text záznamu

Report

Understanding Masked Image Modeling via Learning Occlusion Invariant Feature

Autor: Kong, Xiangwen, Zhang, Xiangyu

Recently, Masked Image Modeling (MIM) achieves great success in self-supervised visual recognition. However, as a reconstruction-based framework, it is still an open question to understand how MIM works, since MIM appears very different from previous

Externí odkaz: http://arxiv.org/abs/2208.04164

Zobrazit plný text záznamu

Report

Revisiting the Critical Factors of Augmentation-Invariant Representation Learning

Autor: Huang, Junqiang, Kong, Xiangwen, Zhang, Xiangyu

We focus on better understanding the critical factors of augmentation-invariant representation learning. We revisit MoCo v2 and BYOL and try to prove the authenticity of the following assumption: different frameworks bring about representations of di

Externí odkaz: http://arxiv.org/abs/2208.00275

Zobrazit plný text záznamu