TheVariational

The topic of generative AI unites key concepts in machine learning. Follows us to learn more about all things AI.

[Ke Li 🍁 @KL_Div]

Diffusion and flow matching-based robot planners are slow and generate noisy and jerky trajectories.

Delighted to share our ICRA 2026 paper, which leverages IMLE to improve planning frequency 19-fold from 4.3 Hz to 83 Hz and reduces jerk by 38% relative to flow matching.

Joint work w/ Grayson Lee, Minh Bui, Shuzi Zhou, Yankai Li and Mo Chen.

(1/7)

[Ethan Caballero @ethanCaballero]

New paper:
We present a "Unified Neural Scaling Law" functional form that accurately models & extrapolates the multivariate scaling behaviors of artificial neural networks as the variables listed in this attached video are varied.
(1/N)

[Peter Pao-Huang @peterpaohuang]

Introducing Flux Matching, a generative modeling paradigm that generalizes diffusion models to vector fields that need not be the score function.

Enables structural priors in the dynamics, faster sampling, interpretable generation, and more!

w/ @StefanoErmon @Xiaojie_Qiu 🧵⤵️

[Michael Albergo @msalbergo]

New paper! Presenting Discrete Flow Maps:

paper: https://t.co/f1RmZry2by
blog: https://t.co/Cnwgf4moY0

A laughable problem for me these days is that @nmboffi and I share a research brain, and we have had, time and again, a conversation that ends with “ha so I guess we’re writing the same paper.” Soon we will return to just doing it together :). Here we are doing it again with discrete flow maps and flow language models! A complete and thorough paper led by @PPotaptchik @json_yim @adhisarav @peholderrieth. We took a bit of time to post it to ensure we understood a few more things about the stability of the loss functions.

Like @osclsd , @FEijkelboom, and @nmboffi , we think this could be a very helpful paradigm for thinking about fast inference and even better alignment!

Here’s our version of the story, and I hope it makes clear how green field this research direction is — we provide a comprehensive picture of the KL losses you can write from the properties of the flow map, some nice geometric proofs about the mean denoiser and the simplex, and find that at this time, the ESD can actually be the most performant, with some caveats. Excited for everyone to work together and push this class of models to their limit!

[TheVariational @TheVariational]

@wellingmax @wellingmax congrats, an amazing set of topics

[TheVariational @TheVariational]

If you don't know @ilyasut @msalbergo @robrombach Now you know! The Continuous Normalizing Flow model is introduced

[TheVariational @TheVariational]

Did you ask for more flow layers!? The continuous normalizing flow @RickyTQChen @peholderrieth @shaulneta is introduced

[TheVariational @TheVariational]

Normalizing flows just got better @laurent_dinh @jaschasd @thoma_gu here’s how, an improvement over linear transforms is discussed

[TheVariational @TheVariational]

Back to basics @alec_helbling @asuzuki1112 back to the flow

[TheVariational @TheVariational]

The blueprint behind normalizing flows @FelixDrRelax @mathusmassias is outlined!

[TheVariational @TheVariational]

A quick look @DaniloJRezende @shakir_za @marcusabrubaker at the fundamentals of normalizing flows

[TheVariational @TheVariational]

@elonmusk @hyhieu226 @NiJinjie latent variable models are making a comeback, https://t.co/V8mEF2MPjr

[TheVariational @TheVariational]

@JCJesseLai @DrYangSong @gimdong58085414 @mittu1204 @StefanoErmon @JCJesseLai what a great body of work, The Variational Book https://t.co/V8mEF2Nn8Z will be out soon, featuring 140+ intuitive illustrations covering all these seminal topics!

[TheVariational @TheVariational]

@JCJesseLai @DrYangSong @gimdong58085414 @mittu1204 @StefanoErmon @JCJesseLai what a great body of work, The Variational Book https://t.co/V8mEF2Nn8Z will be out soon, featuring 140+ intuitive illustrations covering all these seminal topics!

[TheVariational @TheVariational]

@elonmusk @venturetwins @elonmusk unlock the mystery of the "model" see @TheVariational happy to share :)

[TheVariational @TheVariational]

@thimabru Hi Thiago, very much appreciate the enthusiasm, and thank you for asking. Right now I'm editing, refining and aiming for release in the next few months. In the interim, happy to help as can

[TheVariational @TheVariational]

A history of equations @MLStreetTalk @ecsquendor thank you MLST for hosting a session on all things generative AI. check out the 20 page pdf! https://t.co/Mvrd7bCYLq