Xiangming Meng

Assistant Professor @ ZJU-UIUC Institute, Zhejiang University, previous Project Assistant Professor @UTokyo, postdocs @UTokyo @RIKEN_AIP_EN

[Xiangming Meng @XiangmingMeng]

More context: existing training-free samplers usually decide commitments token by token. CLAD instead groups adjacent high-confidence positions into confidence-induced clusters, then uses self-attention from the same forward pass to choose mutually compatible clusters.

[Xiangming Meng @XiangmingMeng]

Glad to share our another new paper CLAD!
CLAD is a training-free decoder for masked diffusion LMs. It commits reliable span-level clusters rather than individual tokens, with attention-guided selection.

Paper: https://t.co/fJ1Zv0frej
Code: https://t.co/gC3nnV70Ry

[Xiangming Meng @XiangmingMeng]

A bit more context: in parallel decoding, committing multiple high-confidence tokens at once can still be redundant if they attend to similar visual regions. VRCD uses token-to-image attention to reduce such overlap during decoding, without extra training or model changes.

[Xiangming Meng @XiangmingMeng]

Glad to share our new paper! We study visual redundancy in parallel decoding for diffusion-based multimodal LLMs and propose VRCD, a lightweight training-free decoding method.

Paper: https://t.co/Yi72e3jQYi
Code: https://t.co/e6LvMl0m7B

[Xiangming Meng @XiangmingMeng]

@ZoubinGhahrama1 Thank you very much Zoubin for your wonderful talk! Is it possible to share your slides of this talk? I did not find it. Many thanks!

[Emtiyaz Khan @EmtiyazKhan]

Not a news to many, but I will be starting as a professor (W3) at TU Darmstadt in June 2026 and setup another branch of our Adaptive Intelligence group.

Also, I am hiring! Please share!

4 PhD positions https://t.co/6uN6pt5gQe
2 post-doc positions https://t.co/6uN6pt5gQe

[Xiangming Meng @XiangmingMeng]

@EmtiyazKhan Congratulations Emti!

[Yang Song @DrYangSong]

Applications change, but the principles are enduring. After a year's hard work led by @JCJesseLai, we are really excited to share this deep, systematic dive into the mathematical principles of diffusion models. This is a monograph we always wished we had.

[Xiangming Meng @XiangmingMeng]

I finally created my blog with first post on one of my earlier work on message passing algorithms. Specifically, we provide a very concise and general derivation of the famous GAMP algorithm from the EP perspective. Your comments are appreciated!

https://t.co/2ctK8aRsHA

[Xiangming Meng @XiangmingMeng]

@FEijkelboom @GrigoryBartosh @wellingmax @canaesseth @jwvdm Thank you for your wonderful thread on VFM. I am really interested in it. May I ask when will you share the code base for VFM? I cannot find it at present. Thanks!

[Xiangming Meng @XiangmingMeng]

@TakaAomidoro Thank you！

[Xiangming Meng @XiangmingMeng]

Our DMPS paper has won the 16th Asian Conference on Machine Learning #ACML Best Paper Runner-Up ward! Many thanks to the ACML 2024 conference committee.

[Xiangming Meng @XiangmingMeng]

DMPS https://t.co/QwraBUhq26 got accepted by ACML2024, where we proposed a simple and fast method to approximate the likelihood score for solving inverse problems using diffusion models. Moreover, a unified approach for both diffusion and flow-based models is illustrated.

[Lenka Zdeborova @zdeborova]

The long-awaited collection of lecture notes from the Summer School on Statistical Physics & Machine Learning, Les Houches 2022, is now published in JStatMech https://t.co/V8tzeimEJB . I am particularly proud of the works the school inspired; see section 3 of the editorial.

[Xiangming Meng @XiangmingMeng]

DMPS https://t.co/QwraBUhq26 got accepted by ACML2024, where we proposed a simple and fast method to approximate the likelihood score for solving inverse problems using diffusion models. Moreover, a unified approach for both diffusion and flow-based models is illustrated.

[Emtiyaz Khan @EmtiyazKhan]

We don't expect Bayesian methods to do so well at large scale, but we can now get decent improvements with variational learning to GPT-2. I wrote a blog about this (first one in a long time). Check it out!
https://t.co/c7ftgBol2x

Paper: https://t.co/GUFi1br9av

A thread below.

[Jeremy Howard @jeremyphoward]

Our new course, "From Deep Learning Foundations to Stable Diffusion", is finally done after 8 months of work!!!

With >30 hours of video content (all free, no ads!), you'll learn how to create and train a Stable Diffusion model starting from pure Python 🧵
https://t.co/bK0PSIzFww

[Emtiyaz Khan @EmtiyazKhan]

The call for papers for our #ICML2023 workshop on "duality principles for modern ML" are now out.

Webpage: https://t.co/SVv3XYsOBp
Deadline: May 22, 2023, AOE

Help us spread the word! Hope many will participate.

@ZeldaMariet @mblondel_ml @tmoellenhoff @BachFrancis @optiML

[Xiangming Meng @XiangmingMeng]

@PierreAlquier @EmtiyazKhan @tmoellenhoff @jyscardioid Thank you Pierre!!

[Xiangming Meng @XiangmingMeng]

@SecPerkinsStan @EmtiyazKhan Thank you a lot!

[Xiangming Meng @XiangmingMeng]

@EmtiyazKhan He is so cute! Congratulations!