Yanzuo Lu (Oliver Lu)

Oct 2026	We release StreamRender-H3, which turns a world written in code into photorealistic video as you play. Built on MiniMax-H3 Ref2VA, it generates each 333–375 ms chunk in about 260 ms on 8×GB200, faster than playback. The code is publicly released.
Sep 2026	RAVEN is accepted by NeurIPS 2026. See you in Sydney!
Sep 2026	I join NVIDIA Research as a research intern, working with Dr. Enze Xie and Prof. Song Han in the Efficient AI Team on the efficiency of generative models.
May 2026	We propose RAVEN, a training-time test framework for distilling real-time autoregressive video generation, together with CM-GRPO, which brings online RL to consistency-model sampling. The code and models are publicly released.
Oct 2025	I start my PhD at Imperial College London supervised by Dr. Jiankang Deng with a fully-funded doctoral scholarship. My research focuses on real-time and long-video world models.
Sep 2025	I leave ByteDance Seed Team with sincere thanks to my close collaborators over the past nearly two years (Weifeng Chen, Huafeng Kuang, Yuxi Ren, Xin Xia, Jie Wu & Xing Wang, etc.), and to my leader Xuefeng Xiao for the valuable opportunity.
Sep 2025	We migrate the acceleration recipes from Seedream models to Bagel, a unified model for understanding and generation, further refining the Hyper-SD methodology into Hyper-Bagel.
Sep 2025	As a core contributor, I work on the acceleration of Seedream 4.0, which is fully rolled out on ByteDance platforms including Doubao alongside the release of its technical report.
Jul 2025	We make an attempt in score distillation by combining it with adversarial training objectives, and our technical report DMDX is accepted to ICCV 2025 as a Highlight.
Jun 2025	I graduate with my MEng from Sun Yat-Sen University and convert to a full-time employee at ByteDance Seed Team.
Apr 2025	I contribute to the acceleration of Seedream 3.0, which is fully rolled out on ByteDance platforms including Doubao, with its technical report released.
Sep 2024	Hyper-SD is accepted to NeurIPS 2024.
Aug 2024	I take charge of training the FLUX and SD3 acceleration LoRAs as an extension of Hyper-SD, featured again on the Hugging Face Trending list.
Jul 2024	Our diffusion-based image editing acceleration work ByteEdit, where I serve as co-first author and use adversarial training and reinforcement learning, is accepted to ECCV 2024.
Apr 2024	We release Hyper-SD, a new state-of-the-art diffusion model acceleration technique trained on SD1.5 and SDXL, featured on the Hugging Face Trending list.
Apr 2024	My first work on diffusion models, CFLD for pose transfer, is accepted to CVPR 2024 as a Highlight. Both the model and code are publicly released.
Dec 2023	I join ByteDance as a research intern, working on diffusion model acceleration.

	RAVEN: Real-time Autoregressive Video Extrapolation with Consistency-model GRPO Yanzuo Lu, Ronglai Zuo, Jiankang Deng Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026 [arXiv] [Project Page] [GitHub] [HuggingFace]
	Adversarial Distribution Matching for Diffusion Distillation Towards Efficient Image and Video Synthesis Yanzuo Lu, Yuxi Ren, Xin Xia, Shanchuan Lin, Xing Wang, Xuefeng Xiao, Andy J Ma, Xiaohua Xie and Jianhuang Lai International Conference on Computer Vision (ICCV), 2025 (Highlight) [arXiv] [Publication]
	Coarse-to-Fine Latent Diffusion for Pose-Guided Person Image Synthesis Yanzuo Lu, Manlin Zhang, Andy J Ma, Xiaohua Xie and Jianhuang Lai IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024 (Highlight) [arXiv] [Publication] [GitHub] [Talk]
	MLNet: Mutual Learning Network with Neighborhood Invariance for Universal Domain Adaptation Yanzuo Lu, Meng Shen, Andy J Ma, Xiaohua Xie and Jianhuang Lai 38th Annual AAAI Conference on Artificial Intelligence (AAAI), 2024 [arXiv] [Publication] [GitHub]
	Hyper-SD: Trajectory Segmented Consistency Model for Efficient Image Synthesis Yuxi Ren, Xin Xia, Yanzuo Lu, Jiacheng Zhang, Jie Wu, Pan Xie, Xing Wang and Xuefeng Xiao Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS), 2024 [arXiv] [Publication] [Project Page] [HuggingFace (4.5M+ downloads, #5 on daily Trending)] [PR]
	ByteEdit: Boost, Comply and Accelerate Generative Image Editing Yuxi Ren, Jie Wu, Yanzuo Lu, Huafeng Kuang, Xin Xia, Xionghui Wang, Qianqian Wang, Yixing Zhu, Pan Xie, Shiyin Wang, Xuefeng Xiao, Yitong Wang, Min Zheng and Lean Fu 18th European Conference on Computer Vision (ECCV), 2024 [arXiv] [Publication] [Project Page]

	Seedream 4.0: Toward Next-generation Multimodal Image Generation Yanzuo Lu (Core Contributor), as part of ByteDance Seed Team arXiv preprint arXiv:2509.20427, 2025 [arXiv] [Project Page]
	Seedream 3.0 Technical Report Yanzuo Lu (Contributor), as part of ByteDance Seed Team arXiv preprint arXiv:2504.11346, 2025 [arXiv] [Project Page]