Julia Linhart

Postdoctoral Researcher @NYUDataScience. AI4Science | Machine Learning | Probabilistic Modeling |Generative Modeling | SBI | Validation & Calibration

[Julia Linhart @jujulinhart]

📄 Paper: https://t.co/zxkgj15awf
💻 Code: https://t.co/GzglhzP68Y
🛠️ Tutorial: https://t.co/l3oE0PevN1

📍ICML Poster: Hall A #3405 - Wed Jul 8, 5-7pm

[Julia Linhart @jujulinhart]

Just presented our work "Diffusion Posterior Sampling for SBI in tall data settings" at #ISBA2026 in Nagoya🇯🇵

➡️ Next stop: #ICML2026 in Seoul 🇰🇷 (Poster #3405). Come say hi!

🙏🏼 Big thanks to my co-authors Gabriel V. Cardoso, Sylvain Le Corff, @plc_rodrigues and @agramfort !

[Julia Linhart @jujulinhart]

💡 Compose diffusion scores correctly and make posterior inference scalable to many i.i.d. observations!

We derive a tractable score composition that enables the efficient sampling of posteriors for large datasets — one network, no augmented datasets, no retraining, no MCMC!

[Julia Kempe @KempeLab]

1/ #1stProof. Our second installment — this time tackling Problem 3, with @scottnarmstrong and @MunosRemi
Also check out our takeaways — and a short “Humor from your bot” interlude — below.

[Shobhita Sundaram @shobsund]

Can a model learn to break its own reasoning plateau?

In our new paper, we show that LLMs can be taught with meta-RL to generate their own "stepping stones" that kickstart learning on hard math problems (0/128 success rate) where direct RL fails.

Paper 📝: https://t.co/lUlrJt6bwq
Blog post 🌐: https://t.co/v1y24h1fP4

(1/n)

[Kyle Cranmer @KyleCranmer]

Incredible turnout and impressive set of posters at the Machine Learning and Physical Sciences workshop
#ML4PS2025
https://t.co/QD6etDojRl

[Shirley Ho @cosmo_shirley]

On a plane to San Diego for Neurips! And @PolymathicAI is presenting 3 conference papers + 4 workshop papers!

From ~0C/35F to 15C/68F + ☀️.
+ Surrounded by my favorite kind of people!

Can't be more excited :)

[Julia Kempe @KempeLab]

I will be recruiting 1-2 PhD students at @NYUDataScience or @NYUCourant CS to work on Machine Learning & applications in NYU's vibrant top ML ecosystem. Check Google Scholar to see our latest research interests. Interested? Please mention my name in your application. Deadl. 12/12

[Shirley Ho @cosmo_shirley]

Extremely proud and honored to be named among the great minds of @schmidtsciences AI2050 Senior fellows!

A special shout-out to my team @PolymathicAI!
For building out amazing foundation models for the sciences!

HUGE thanks to my team at @SimonsFdn and my mentors @DavidSpergel and @ericschmidt for believing in my vision and supporting me all the way through!

This won't be possible without you! 🥰

[Kyle Cranmer @KyleCranmer]

Is scale all you need? Or is there still a role for incorporating domain knowledge and inductive bias? While I was in Heidelberg, I took some time to write a short essay on this question called "The Bittersweet Lesson".
https://t.co/DQEItqXomF #HLF25

[Julia Linhart @jujulinhart]

✨ Excited to begin my Postdoc at @NYUDataScience, working with Shirley Ho @cosmo_shirley, Julia Kempe @KempeLab and Uroš Seljak on deep generative modeling and Bayesian inference for cosmology 🚀

[Antoine Wehenkel @WehenkelAntoine]

Alert non-LLM internship position in Zurich Health AI 🍏🚨🧑‍💻
If you 1) have a strong background in deep probabilistic modeling, SBI, hybrid learning or causality 2) are curious about exciting health applications, you should apply!
https://t.co/324qlCAtPj

[Julia Linhart @jujulinhart]

I'm so proud of our @sbi_devs community! This new release brings a flexible API, accessible tutorials, and solid validation methods—making `sbi` more practical and reliable for everyone. Big thanks to @mackelab for making this happen!

[sbi developers @sbi_devs]

We just released a new version of `sbi`, and this one has _a ton_ of new features! Many of these features are thanks to more than 30 (mostly new) contributors. We are very excited about the growing community and the new release! 🧵 1/8

[sbi developers @sbi_devs]

We’re excited to share that the SBI package is now officially @NumFOCUS affiliated and growing into a community project. A lot of upcoming release originated from a hackathon earlier this year, and there are more events planned soon. Stay tuned for details! #opensource #hackathon

[Florentin Guth @FlorentinGuth]

Ideas can have long-term impact, while SOTA is now already obsolete by the time of the conference. Yet most reviewers (and sometimes ACs) only care about tables of numbers and scale. Shouldn't we discuss what we learned from the paper? Good engineering requires good science.

[Gilles Louppe @glouppe]

Watching PhD students lose their last crumbs of illusion about science when faced with stubborn, unreasonable, or completely absent reviewers is heartbreaking. We've got to do better as a community. #NeurIPS2024

[Benjamin Kurt Miller @bkmi13]

Nice to see some attention for SBI @ylecun!

... especially on the day when I turned in my PhD manuscript about the topic!

[Umberto Picchini @uPicchini]

new paper on simulation-based inference using Gaussian mixtures to learn likelihood and posterior, without neural networks 😱😱😱

Our SeMPLE (Sequential Mixture Posterior and Likelihood Estimation) provides excellent inference and excels when the posterior is multimodal. 1/n

[Moreau Thomas @tomamoral]

@OwkinScience perfectly captured the atmosphere of our benchopt sprint !
https://t.co/y1oOoy2S8T

A lot of exchanges, coding, and fun! 😄

So proud of what we achieved during these three days!
Here is a recap of this sprint's output: https://t.co/3dhKOQtbae

[Moreau Thomas @tomamoral]

We just finished a very productive benchopt sprint, gathering 35 participants over 3 days, hosted at @Owkin_Science, Google AI Hub & SCAI @Sorbonne_Univ.
We merged 11 PRs, updated 9 benchmarks, and developed 7 new ones.
Here are some key highlights!

https://t.co/ExbizNMBQC