msalbergo's profile picture. Junior fellow at the Society of Fellows at @Harvard and @iaifi_news fellow, incoming Assistant Professor at @Harvard and the @KempnerInst views my own

Michael Albergo

@msalbergo

Junior fellow at the Society of Fellows at @Harvard and @iaifi_news fellow, incoming Assistant Professor at @Harvard and the @KempnerInst views my own

Épinglé

📍🐳🐳🐳📍New paper with an awesome crew! We expand the design space of any-order masked diffusion models to rigorously make possible variable length generation (and potentially more fun things to come) via joint interpolants. Solid performance benefits! 👇👇👇 Super proud of…

Announcing Flexible Masked Diffusion Models (FlexMDMs)—a new diffusion language model for flexible-length sequences. 🚨 Solves MDMs' fixed-length issue + retrains any-order sampling 🚨 <1000 GPU-hrs to fine-tune LLaDA-8B into FlexMDM (GSM8K 58→67%, HumanEval-infill: 52→65%)



Michael Albergo a reposté

New work: “GLASS Flows: Transition Sampling for Alignment of Flow and Diffusion Models”. GLASS generates images by sampling stochastic Markov transitions with ODEs - allowing us to boost text-image alignment for large-scale models at inference time! arxiv.org/pdf/2509.25170 [1/7]


Michael Albergo a reposté

Diffusion models are powerful 🔥, but adapting them at inference time without retraining is challenging ⁉️. We introduce DriftLite, a lightweight 🪶, training-free 😀, inference-time scaling method that actively steers inference and absorbs instability, preventing weight…


Michael Albergo a reposté

Neat framework to unify and understand consistency-like models by @nmboffi, @msalbergo and Eric Vanden-Eijnden !

Consistency models, CTMs, shortcut models, align your flow, mean flow... What's the connection, and how should you learn them in practice? We show they're all different sides of the same coin connected by one central object: the flow map. arxiv.org/abs/2505.18825 🧵(1/n)



We've cleaned up the story big time on flow maps. Check out @nmboffi's slick repo implementing all the many ways to go about them, and stay tuned for a bigger release 🤠 arxiv.org/pdf/2505.18825 flow-maps.github.io

Consistency models, CTMs, shortcut models, align your flow, mean flow... What's the connection, and how should you learn them in practice? We show they're all different sides of the same coin connected by one central object: the flow map. arxiv.org/abs/2505.18825 🧵(1/n)



Michael Albergo a reposté

📢Thrilled to share that I'll be joining Harvard and the Kempner Institute as an Assistant Professor starting Fall 2026! I'll be recruiting students this year for the Fall 2026 admissions cycle. Hope you apply!


Our paper is now out in PRX! Great work by @aavishkar_patel and Peter

This study was conducted by @aavishkar_patel at @FlatironInst, Peter Lunts at @Harvard, and @msalbergo at @nyuniversity.



Michael Albergo a reposté

🚨 Deadline Extension! 🚨 The FPI workshop submission deadline is now August 29th. Submit your latest research on Probabilistic Inference and Sampling. Don't miss out! fpineurips.framer.website #NeurIPS2025 #ProbabilisticInference #AI

AlexanderTong7's tweet image. 🚨 Deadline Extension! 🚨

The FPI workshop submission deadline is now August 29th.

Submit your latest research on Probabilistic Inference and Sampling. Don&apos;t miss out!

fpineurips.framer.website
#NeurIPS2025 #ProbabilisticInference #AI

Michael Albergo a reposté

Working on Probabilistic Inference and Sampling 🤔? Consider submitting to the FPI workshop at #NeurIPS2025 this year. 🔗fpineurips.framer.website 🚨Deadline is coming up, and it's Aug 22nd! So submit your best work, early!


Computer scientist friends: nominate your students/postdocs for a junior fellowship @Harvard: socfell.fas.harvard.edu it's a wonderful interdisciplinary experience!


Michael Albergo a reposté

We’re thrilled to introduce the 2025 cohort of #KempnerInstitute Graduate Fellows! This year’s recipients include grad students enrolled across five @Harvard Ph.D. programs. Read more: bit.ly/3TYzZ5H @hseas, @HarvardGSAS, @harvardphysics @PiN_Harvard #AI #NeuroAI #ML


Michael Albergo a reposté

Pitch the dataset that could spark the next AI for Science revolution 🚀 The PDB revolutionized structural biology (and even helped win a🏅Nobel Prize in 2024). We’re hunting for the next breakthrough dataset that could unlock similar leaps across science—and we want your idea!

AI_for_Science's tweet image. Pitch the dataset that could spark the next AI for Science revolution 🚀
The PDB revolutionized structural biology (and even helped win a🏅Nobel Prize in 2024). We’re hunting for the next breakthrough dataset that could unlock similar leaps across science—and we want your idea!

Michael Albergo a reposté

AI for Science will be returning to @NeurIPSConf 2025! We aim to bring together scientists and AI researchers to discuss the reach and limits of AI for Scientific Discovery 🚀 📖 Workshop submission deadline: Aug 22 💡 Dataset proposal competition: more details coming soon

AI_for_Science's tweet image. AI for Science will be returning to @NeurIPSConf 2025! We aim to bring together scientists and AI researchers to discuss the reach and limits of AI for Scientific Discovery 🚀

📖 Workshop submission deadline: Aug 22
💡 Dataset proposal competition: more details coming soon
AI_for_Science's tweet image. AI for Science will be returning to @NeurIPSConf 2025! We aim to bring together scientists and AI researchers to discuss the reach and limits of AI for Scientific Discovery 🚀

📖 Workshop submission deadline: Aug 22
💡 Dataset proposal competition: more details coming soon

Michael Albergo a reposté

1/ Where do Probabilistic Models, Sampling, Deep Learning, and Natural Sciences meet? 🤔 The workshop we’re organizing at #NeurIPS2025! 📢 FPI@NeurIPS 2025: Frontiers in Probabilistic Inference – Learning meets Sampling Learn more and submit → fpiworkshop.org


Michael Albergo a reposté

A gentle reminder that TMLR is a great journal that allows you to submit your papers when they are ready rather than rushing to meet conference deadlines. The review process is fast, there are no artificial acceptance rates, and you have more space to present your ideas in the…


Dear @NeurIPSConf -- it seems OpenReview is down entirely, and we cannot submit reviews for the upcoming review deadline tonight. Please share if you are having a similar issue. #neurips2025


Michael Albergo a reposté

1/6 Infinite-dim SGD in linear regression is the strawman model for studying scaling laws, critical batch sizes, and LR schedules. We revisit (and simplify) its analysis using just linear algebra, making it easier to derive and reason about. No PSD operators. No tensor calculus.


Loading...

Something went wrong.


Something went wrong.