A Hidden Semantic Bottleneck in Conditional Embeddings of Diffusion Transformers.

Kang Zhang 0008 Korea Advanced Institute of Science and Technology, South Korea Harbin Institute of Technology, Department of Microwave Engineering, China https://scholar.google.com/citations?user=nj19btQAAAAJ https://openreview.net/profile?id=~Kang_Zhang6 https://orcid.org/0000-0003-2761-9383 https://www.wikidata.org/entity/Q104749114 Kang Zhang Kang Zhang 0001 http://utd.edu/~kzhang/ https://dl.acm.org/profile/81350576955 https://orcid.org/0000-0003-3802-7535 https://www.wikidata.org/entity/Q89560574 https://d-nb.info/gnd/1185436634 https://id.loc.gov/authorities/names/nb2003067129 https://mathgenealogy.org/id.php?id=127395 https://zbmath.org/authors/?q=ai:zhang.kang.1 Hong Kong University of Science and Technology, Guangzhou, China University of Texas at Dallas, TX, USA Kang Zhang 0002 Louisiana State University, Baton Rouge, LA, USA https://orcid.org/0000-0003-0576-6980 Kang Zhang 0003 Hohai University, Nanjing, China https://orcid.org/0000-0002-4594-5599 https://ieeexplore.ieee.org/author/37088411869 Kang Zhang 0004 Tsinghua University, Beijing, China Kang Zhang 0005 JD.com, Beijing, China Kang Zhang 0006 Chengde Petroleum College, Chengde, China Kang Zhang 0007 Ocean University of China, Qingdao, China

Trung X. Pham Kang Zhang 0008 Ji Woo Hong Chang D. Yoo A Hidden Semantic Bottleneck in Conditional Embeddings of Diffusion Transformers. 2026 February abs/2602.21596 CoRR https://doi.org/10.48550/arXiv.2602.21596 db/journals/corr/corr2602.html#abs-2602-21596 streams/journals/corr

Axi Niu Kang Zhang 0008 Qingsen Yan Hao Jin Jinqiu Sun Yanning Zhang 0001 DualTSR: Unified Dual-Diffusion Transformer for Scene Text Image Super-Resolution. 2026 March abs/2603.14207 CoRR https://doi.org/10.48550/arXiv.2603.14207 db/journals/corr/corr2603.html#abs-2603-14207 streams/journals/corr

Kang Zhang 0008 Suyeon Lee Arda Senocak Joon Son Chung Cinematic Audio Source Separation Using Visual Cues. 2026 March abs/2603.26113 CoRR https://doi.org/10.48550/arXiv.2603.26113 db/journals/corr/corr2603.html#abs-2603-26113 streams/journals/corr

Axi Niu Kang Zhang 0008 Trung Xuan Pham Pei Wang Jinqiu Sun In So Kweon Yanning Zhang 0001 Learning From Multi-Perception Features for Real-Word Image Super-Resolution. 6535-6548 2025 July 35 IEEE Trans. Circuits Syst. Video Technol. 7 https://doi.org/10.1109/TCSVT.2024.3374967 db/journals/tcsv/tcsv35.html#NiuZPWSKZ25 streams/journals/tcsv

Kang Zhang 0008 Trung X. Pham Suyeon Lee Axi Niu Arda Senocak Joon Son Chung Model-Guided Dual-Role Alignment for High-Fidelity Open-Domain Video-to-Audio Generation. 2025 NeurIPS http://papers.nips.cc/paper_files/paper/2025/hash/33a0db5c55008b14cc088fb086612b1a-Abstract-Conference.html conf/nips/2025 db/conf/nips/neurips2025.html#ZhangPLNSC25 Chenshuang Zhang Kang Zhang 0008 Joon Son Chung In So Kweon Junmo Kim 0002 Chengzhi Mao Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision. 2025 NeurIPS http://papers.nips.cc/paper_files/paper/2025/hash/396446770f5e8496ca1feb02079d4fb7-Abstract-Conference.html conf/nips/2025 db/conf/nips/neurips2025.html#ZhangZCKKM25

Trung X. Pham Kang Zhang 0008 Ji Woo Hong Xuran Zheng Chang D. Yoo E-MD3C: Taming Masked Diffusion Transformers for Efficient Zero-Shot Object Customization. 2025 February abs/2502.09164 CoRR https://doi.org/10.48550/arXiv.2502.09164 db/journals/corr/corr2502.html#abs-2502-09164 streams/journals/corr

Kang Zhang 0008 Trung X. Pham Suyeon Lee Axi Niu Arda Senocak Joon Son Chung Model-Guided Dual-Role Alignment for High-Fidelity Open-Domain Video-to-Audio Generation. 2025 October abs/2510.24103 CoRR https://doi.org/10.48550/arXiv.2510.24103 db/journals/corr/corr2510.html#abs-2510-24103 streams/journals/corr

Chenshuang Zhang Kang Zhang 0008 Joon Son Chung In So Kweon Junmo Kim 0002 Chengzhi Mao Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision. 2025 December abs/2512.02339 CoRR https://doi.org/10.48550/arXiv.2512.02339 db/journals/corr/corr2512.html#abs-2512-02339 streams/journals/corr

Axi Niu Trung X. Pham Kang Zhang 0008 Jinqiu Sun Yu Zhu 0004 Qingsen Yan In So Kweon Yanning Zhang 0001 ACDMSR: Accelerated Conditional Diffusion Models for Single Image Super-Resolution. 492-504 2024 June 70 IEEE Trans. Broadcast. 2 https://doi.org/10.1109/TBC.2024.3374122 db/journals/tbc/tbc70.html#NiuPZSZYKZ24

Joshua Tian Jin Tee Kang Zhang 0008 Hee Suk Yoon Dhananjaya N. Gowda Chanwoo Kim 0001 Chang D. Yoo Physics Informed Distillation for Diffusion Models. 2024 2024 Trans. Mach. Learn. Res. https://openreview.net/forum?id=rOvaUsF996 db/journals/tmlr/tmlr2024.html#Tee0YG0Y24

Chenshuang Zhang Chaoning Zhang Kang Zhang 0008 Axi Niu Junmo Kim 0002 In So Kweon Towards Understanding Dual BN In Hybrid Adversarial Training. 2024 2024 Trans. Mach. Learn. Res. https://openreview.net/forum?id=bQKHMSE4SH db/journals/tmlr/tmlr2024.html#ZhangZ0N0K24

Hee Suk Yoon Eunseop Yoon Joshua Tian Jin Tee Kang Zhang 0008 Yu-Jung Heo Du-Seong Chang Chang D. Yoo BI-MDRG: Bridging Image History in Multimodal Dialogue Response Generation. 378-396 2024 ECCV (31) https://doi.org/10.1007/978-3-031-72751-1_22 conf/eccv/2024-31 db/conf/eccv/eccv2024-31.html#YoonYTZHCY24 streams/conf/eccv Trung X. Pham Kang Zhang 0008 Chang D. Yoo Cross-view Masked Diffusion Transformers for Person Image Synthesis. 2024 ICML https://proceedings.mlr.press/v235/pham24b.html https://openreview.net/forum?id=jEoIkNkqyc conf/icml/2024 db/conf/icml/icml2024.html#Pham0Y24 40611-40641

Trung X. Pham Kang Zhang 0008 Chang D. Yoo Cross-view Masked Diffusion Transformers for Person Image Synthesis. 2024 abs/2402.01516 CoRR https://doi.org/10.48550/arXiv.2402.01516 db/journals/corr/corr2402.html#abs-2402-01516

Chenshuang Zhang Chaoning Zhang Kang Zhang 0008 Axi Niu Junmo Kim 0002 In So Kweon Towards Understanding Dual BN In Hybrid Adversarial Training. 2024 abs/2403.19150 CoRR https://doi.org/10.48550/arXiv.2403.19150 db/journals/corr/corr2403.html#abs-2403-19150

Hee Suk Yoon Eunseop Yoon Joshua Tian Jin Tee Kang Zhang 0008 Yu-Jung Heo Du-Seong Chang Chang D. Yoo BI-MDRG: Bridging Image History in Multimodal Dialogue Response Generation. 2024 abs/2408.05926 CoRR https://doi.org/10.48550/arXiv.2408.05926 db/journals/corr/corr2408.html#abs-2408-05926 streams/journals/corr

Joshua Tian Jin Tee Kang Zhang 0008 Hee Suk Yoon Dhananjaya N. Gowda Chanwoo Kim 0001 Chang D. Yoo Physics Informed Distillation for Diffusion Models. 2024 abs/2411.08378 CoRR https://doi.org/10.48550/arXiv.2411.08378 db/journals/corr/corr2411.html#abs-2411-08378 streams/journals/corr

Trung Xuan Pham Axi Niu Kang Zhang 0008 Joshua Tian Jin Tee Ji Woo Hong Chang D. Yoo Self-Supervised Visual Representation Learning via Residual Momentum. 116706-116720 2023 11 IEEE Access https://doi.org/10.1109/ACCESS.2023.3325842 db/journals/access/access11.html#PhamNZTHY23

Axi Niu Kang Zhang 0008 Trung X. Pham Jinqiu Sun Yu Zhu 0004 In So Kweon Yanning Zhang 0001 CDPMSR: Conditional Diffusion Probabilistic Models for Single Image Super-Resolution. 615-619 2023 ICIP https://doi.org/10.1109/ICIP49359.2023.10222191 conf/icip/2023 db/conf/icip/icip2023.html#NiuZPSZKZ23

Axi Niu Kang Zhang 0008 Trung X. Pham Jinqiu Sun Yu Zhu 0004 In So Kweon Yanning Zhang 0001 CDPMSR: Conditional Diffusion Probabilistic Models for Single Image Super-Resolution. 2023 abs/2302.12831 CoRR https://doi.org/10.48550/arXiv.2302.12831 db/journals/corr/corr2302.html#abs-2302-12831

Axi Niu Kang Zhang 0008 Trung X. Pham Pei Wang Jinqiu Sun In So Kweon Yanning Zhang 0001 Learning from Multi-Perception Features for Real-Word Image Super-resolution. 2023 abs/2305.18547 CoRR https://doi.org/10.48550/arXiv.2305.18547 db/journals/corr/corr2305.html#abs-2305-18547

Axi Niu Trung Xuan Pham Kang Zhang 0008 Jinqiu Sun Yu Zhu 0004 In So Kweon Yanning Zhang 0001 ACDMSR: Accelerated Conditional Diffusion Models for Single Image Super-Resolution. 2023 abs/2307.00781 CoRR https://doi.org/10.48550/arXiv.2307.00781 db/journals/corr/corr2307.html#abs-2307-00781

Axi Niu Kang Zhang 0008 Joshua Tian Jin Tee Trung X. Pham Jinqiu Sun Chang D. Yoo In So Kweon Yanning Zhang 0001 DifAugGAN: A Practical Diffusion-style Data Augmentation for GAN-based Single Image Super-resolution. 2023 abs/2311.18508 CoRR https://doi.org/10.48550/arXiv.2311.18508 db/journals/corr/corr2311.html#abs-2311-18508

Chaoning Zhang Kang Zhang 0008 Trung X. Pham Axi Niu Zhinan Qiao Chang D. Yoo In So Kweon Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo. 14421-14430 2022 CVPR https://doi.org/10.1109/CVPR52688.2022.01404 conf/cvpr/2022 db/conf/cvpr/cvpr2022.html#ZhangZPNQYK22 Chaoning Zhang Philipp Benz Adil Karjauv Jae-Won Cho Kang Zhang 0008 In So Kweon Investigating Top-k White-Box and Transferable Black-box Attack. 15064-15073 2022 CVPR https://doi.org/10.1109/CVPR52688.2022.01466 conf/cvpr/2022 db/conf/cvpr/cvpr2022.html#ZhangBKCZK22 Chaoning Zhang Kang Zhang 0008 Chenshuang Zhang Axi Niu Jiu Feng Chang D. Yoo In So Kweon Decoupled Adversarial Contrastive Learning for Self-supervised Adversarial Robustness. 725-742 2022 ECCV (30) https://doi.org/10.1007/978-3-031-20056-4_42 conf/eccv/2022-30 db/conf/eccv/eccv2022-30.html#ZhangZZNFYK22 Chaoning Zhang Kang Zhang 0008 Chenshuang Zhang Trung X. Pham Chang D. Yoo In So Kweon How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learning. 2022 ICLR https://openreview.net/forum?id=bwq6O4Cwdl conf/iclr/2022 db/conf/iclr/iclr2022.html#ZhangZZPYK22

Chaoning Zhang Kang Zhang 0008 Axi Niu Chenshuang Zhang Jiu Feng Chang D. Yoo In So Kweon Noise Augmentation Is All You Need For FGSM Fast Adversarial Training: Catastrophic Overfitting And Robust Overfitting Require Different Augmentation. 2022 abs/2202.05488 CoRR https://arxiv.org/abs/2202.05488 db/journals/corr/corr2202.html#abs-2202-05488

Chaoning Zhang Kang Zhang 0008 Chenshuang Zhang Trung X. Pham Chang D. Yoo In So Kweon How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learning. 2022 abs/2203.16262 CoRR https://doi.org/10.48550/arXiv.2203.16262 db/journals/corr/corr2203.html#abs-2203-16262

Chaoning Zhang Kang Zhang 0008 Trung X. Pham Axi Niu Zhinan Qiao Chang D. Yoo In So Kweon Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo. 2022 abs/2203.17248 CoRR https://doi.org/10.48550/arXiv.2203.17248 db/journals/corr/corr2203.html#abs-2203-17248

Chaoning Zhang Philipp Benz Adil Karjauv Jae-Won Cho Kang Zhang 0008 In So Kweon Investigating Top-k White-Box and Transferable Black-box Attack. 2022 abs/2204.00089 CoRR https://doi.org/10.48550/arXiv.2204.00089 db/journals/corr/corr2204.html#abs-2204-00089

Agus Gunawan Xu Yin Kang Zhang 0008 Understanding and Improving Group Normalization. 2022 abs/2207.01972 CoRR https://doi.org/10.48550/arXiv.2207.01972 db/journals/corr/corr2207.html#abs-2207-01972

Chaoning Zhang Kang Zhang 0008 Chenshuang Zhang Axi Niu Jiu Feng Chang D. Yoo In So Kweon Decoupled Adversarial Contrastive Learning for Self-supervised Adversarial Robustness. 2022 abs/2207.10899 CoRR https://doi.org/10.48550/arXiv.2207.10899 db/journals/corr/corr2207.html#abs-2207-10899

Chaoning Zhang Chenshuang Zhang Junha Song John Seon Keun Yi Kang Zhang 0008 In So Kweon A Survey on Masked Autoencoder for Self-supervised Learning in Vision and Beyond. 2022 abs/2208.00173 CoRR https://doi.org/10.48550/arXiv.2208.00173 db/journals/corr/corr2208.html#abs-2208-00173

Trung X. Pham Chaoning Zhang Axi Niu Kang Zhang 0008 Chang D. Yoo On the Pros and Cons of Momentum Encoder in Self-Supervised Visual Representation Learning. 2022 abs/2208.05744 CoRR https://doi.org/10.48550/arXiv.2208.05744 db/journals/corr/corr2208.html#abs-2208-05744

Trung X. Pham Axi Niu Kang Zhang 0008 Sultan Rizky Hikmawan Madjid Ji Woo Hong Daehyeok Kim Joshua Tian Jin Tee Chang D. Yoo Self-Supervised Visual Representation Learning via Residual Momentum. 2022 abs/2211.09861 CoRR https://doi.org/10.48550/arXiv.2211.09861 db/journals/corr/corr2211.html#abs-2211-09861

Kang Zhang 0008 Shengchang Lan Guiyuan Zhang On the Effect of Training Convolution Neural Network for Millimeter-Wave Radar-Based Hand Gesture Recognition. 259 2021 21 Sensors 1 https://doi.org/10.3390/s21010259 https://www.wikidata.org/entity/Q104749115 db/journals/sensors/sensors21.html#ZhangLZ21

Philipp Benz Du-Seong Chang Jae-Won Cho Joon Son Chung Jiu Feng Dhananjaya GowdaDhananjaya N. Gowda Agus Gunawan Yu-Jung Heo Ji Woo Hong Hao Jin Adil Karjauv Chanwoo Kim 0001 Daehyeok Kim Junmo Kim 0002 In-So KweonIn So Kweon Shengchang Lan Suyeon Lee Sultan Rizky Hikmawan Madjid Chengzhi Mao Axi Niu Trung X. PhamTrung Xuan Pham Zhinan Qiao Arda Senocak Junha Song Jinqiu Sun Joshua Tian Jin Tee Pei Wang Qingsen Yan John Seon Keun Yi Xu Yin Chang Dong YooChang D. Yoo Eunseop Yoon Hee Suk Yoon Chaoning Zhang Chenshuang Zhang Guiyuan Zhang Yanning Zhang 0001 Xuran Zheng Yu Zhu 0004