基于深度卷积神经网络的 ImageNet 分类
ImageNet Classification with Deep Convolutional Neural Networks
深度卷积网络将数据、GPU 计算与表示学习结合，为现代视觉研究提供了重要起点。
THE RESEARCH ATLAS
Qwen-Image-2.0-RL：图像生成的强化学习与蒸馏
扩散模型后训练流程结合任务奖励、强化学习与同策略蒸馏。
Qwen-Image-2.0-RL Technical Report
为什么值得读
结合奖励设计及评测范围，理解指令遵循与视觉质量的改善。
引用
@misc{xu2026qwen,
title = {Qwen-Image-2.0-RL Technical Report},
author = {Yixian Xu and Kaiyuan Gao and Yuxiang Chen and Yilei Chen and Zecheng Tang and Zihao Liu and Zikai Zhou and Deqing Li and Hao Meng and Kuan Cao and Jiahao Li and Jie Zhang and Liang Peng and Lihan Jiang and Ningyuan Tang and Shengming Yin and Tianhe Wu and Xiaoyue Chen and Yan Shu and Yanran Zhang and Yi Wang and Yu Wu and Yujia Wu and Zekai Zhang and Zhendong Wang and Xiao Xu and Kun Yan and Chenfei Wu},
year = {2026},
eprint = {2606.27608},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.27608}
}论文与代码版权归原作者；引用及许可请以原始来源为准。
相关研究
去噪扩散概率模型
Denoising Diffusion Probabilistic Models
模型学习逐步反转加噪过程，为图像生成建立了重要的扩散建模路线。
从自然语言监督中学习可迁移的视觉模型
Learning Transferable Visual Models From Natural Language Supervision
通过对比学习对齐图像与自然语言描述，使视觉表示能够迁移到新类别。