Linqi (Alex) Zhou
@linqi_zhou
Followers
798
Following
288
Media
7
Statuses
79
Research Scientist @LumaLabsAI. Ph.D. Student at Stanford University (on leave). Prev co-founder @apparatelabs (acq.).
Los Angeles, CA
Joined August 2019
SO excited to finally share my work at Luma! We introduce Inductive Moment Matching, a new generative paradigm that can be trained stably with a single model and single objective from scratch, achieving 1.99 FID on ImageNet-256x256 in 8 steps and 1.98 FID on CIFAR-10 in 2 steps.
Today, we release Inductive Moment Matching (IMM): a new pre-training paradigm breaking the algorithmic ceiling of diffusion models. Higher sample quality. 10x more efficient. Single-stage, single network, stable training. Read more:
7
19
114
RT @baaadas: @linqi_zhou and I will be presenting IMM ( @ ICML on Tuesday 4pm (oral) and 4:30pm-6:00pm (poster).….
lu.ma
Come hang out and have drinks with the amazing Luma AI team at ICML 2025! 🥰🍻
0
2
0
IMM full training code is released at @baaadas and I are presenting the paper (oral) at ICML. If you want to chat, please also join our Happy Hour on Tuesday!.
lu.ma
Come hang out and have drinks with the amazing Luma AI team at ICML 2025! 🥰🍻
SO excited to finally share my work at Luma! We introduce Inductive Moment Matching, a new generative paradigm that can be trained stably with a single model and single objective from scratch, achieving 1.99 FID on ImageNet-256x256 in 8 steps and 1.98 FID on CIFAR-10 in 2 steps.
0
6
38
RT @wanqiao_xu: Thanks for the recognition of the best theory paper award at ICML 2025 EXAIT: Congratulations to t….
0
4
0
RT @chenhao_chao: (1/5) 👑 New Discrete Diffusion Model — MDM-Prime. Why restrict tokens to just masked or unmasked in masked diffusion mode….
0
32
0
RT @allenainie: Decision-making with LLM can be studied with RL! Can an agent solve a task with text feedback (OS terminal, compiler, a per….
0
25
0
RT @meihuadang: #CVPR2025 "Personalized Preference Fine-tuning of Diffusion Models". We extend DPO to align text-to-image diffusion models….
0
55
0
Excited to announce that IMM is accepted as an oral for ICML. As I’ll be going to CVPR as well, if you’d like to chat about research see you at @LumaLabsAI open bar event.
SO excited to finally share my work at Luma! We introduce Inductive Moment Matching, a new generative paradigm that can be trained stably with a single model and single objective from scratch, achieving 1.99 FID on ImageNet-256x256 in 8 steps and 1.98 FID on CIFAR-10 in 2 steps.
1
8
40
RT @gravicle: "Pre-training as we know it will end, Data is not growing". Limited text data is blocking the path to useful general intellig….
0
18
0
Thanks @iScienceLuvr for sharing our latest work. Our method surpasses diffusion and Flow Matching while being trained stably from scratch. Checkout our blog post:
lumalabs.ai
Inductive Moment Matching surpasses diffusion models in speed and sample quality.
Inductive Moment Matching. Luma AI introduces a new class of generative models for one- or few-step sampling with a single-stage training procedure. Surpasses diffusion models on ImageNet-256×256 with 1.99 FID using only 8 inference steps and achieves state-of-the-art 2-step
0
0
18
RT @baaadas: As one of the people who popularized the field of diffusion models, I am excited to share something that might be the “beginni….
0
112
0
RT @LumaLabsAI: Today, we release Inductive Moment Matching (IMM): a new pre-training paradigm breaking the algorithmic ceiling of diffusio….
0
226
0
@StefanoErmon @baaadas Also special thanks to Wanqiao @wanqiao_xu, Sam @_sam_sinha_, and my previous co-founders Bokui @shenbokui and Connor @connorzl for their helpful suggestions and amazing support along the way.
0
1
5
IMM Paper: Github: Position Paper: Huge thanks to Stefano @StefanoErmon and Jiaming @baaadas for this amazing collaboration. If you are interested in the new frontier of advanced generative methods, join us!.
1
1
7
Checkout our blog post for details at: In addition, we release a position paper explaining our motivation and IMM’s broader implications. We hope it can reveal problems of existing methods and open up new design spaces for future generative models to come.
lumalabs.ai
Inductive Moment Matching surpasses diffusion models in speed and sample quality.
1
1
2
RT @Scobleizer: One image is a lot more powerful than I was thinking. Learned in the streets of New York today and here:.
0
12
0
Our avatar livestreams can be found at Go interact with them!.
0
0
0
I want to share about this amazing project I have been working on for the past few months. I could not have done this without the incredible @willbokuishen, @connorzl, @GordonWetzstein , and generous support from our advisors Leo Guibas and @StefanoErmon!.
Introducing Proteus 0.1, REAL-TIME video generation that brings life to your AI. Proteus can laugh, rap, sing, blink, smile, talk, and more. From a single image!. Come meet Proteus on Twitch in real-time. ↓.Sign up for API waitlist: 1/11
4
5
29
Checkout this awesome work by @_Aaditya_Prasad!.
Diffusion Policies are powerful and widely used. We made them much faster. Consistency Policy bridges consistency distillation techniques to the robotics domain and enables 10-100x faster policy inference with comparable performance. Accepted at #RSS2024
0
0
5