Felix Koehler @ NeurIPS 2025
@felix_m_koehler
Followers
2K
Following
360
Media
154
Statuses
385
🤖 Machine Learning & 🌊 Simulation | 📺 YouTuber | 🧑🎓 PhD student @thuereyGroup
Munich
Joined July 2021
Are you interested in Differentiable Programming/Physics & Autodiff? @SimAI4Science is currently hosting the first Tesseract hackathon. Participate, build cool stuff and win prizes! Find more details here: https://t.co/2WWQrdm411 Submission Deadline: Jan 5 2026
pasteurlabs.ai
Build something cool with Tesseract, an open-source software ecosystem for creating end-to-end differentiable pipelines that bridge research and simulation workflows. Win awesome prizes.
0
1
12
Get a quick overview of the paper on our project page:
0
0
0
This work shines light on what the neural emulators learn and (I think) highlights the importance of understanding the surrogate pipeline more holistically.
1
0
0
Under certain conditions, inductive biases and differences between training and testing allow networks to outlearn structured flaws having them become better than their training source.
1
0
0
Please stop by if you are interested in the effect of synthetic training data quality (with respect to numerical errors) on learned neural surrogates/emulators/operators.
1
0
0
Excited to present "Emulator Superiority" today at NeurIPS in San Diego (# 2106, 4:30pm-7:30pm PT).
1
0
4
I will attend #NeurIPS 2025 in San Diego and will spend a few days prior in the Bay Area. DM me if you want to chat 😊. At NeurIPS, I will present my paper "Emulator Superiority": https://t.co/8QaRSfjpIG
3
0
6
Last week, I had the honor to speak about this at the RISEof speaking ML seminar, hosted by Olof Mogren. You can find the recording here
0
0
1
🧵 Project Page: https://t.co/zMV9lfdxxZ 📃 Paper: https://t.co/q4nlrRsKVG Feel free to stop by at NeurIPS in San Diego during the poster session on Friday evening 4:30 p.m. PST — 7:30 p.m. PST at # 2106. 😊 I am looking forward to the exchange!
1
0
1
Proven theoretically for linear PDEs, validated experimentally on nonlinear ones like Burgers' equation. Time to rethink data-driven ML benchmarks for higher physical fidelity!
1
0
0
We show neural emulators trained on low-fidelity data can outperform their source simulators—thanks to inductive biases and smarter error accumulation—beating them against high-fidelity references.
1
0
0
Can your AI surpass the simulator that taught it? What if the key to more accurate PDE modeling lies in questioning your training data's origins? 🤔 Excited to share my #NeurIPS 2025 paper with @thuereyGroup: "Neural Emulator Superiority"!
1
2
9
📢 Calling for (4-page) workshop papers on #differentiable programming and #SciML. This is a great opportunity to showcase your work or work in progress alongside keynote presentations by Petros Koumoutsakos, @PatrickKidger, @thuereyGroup & Astrid Walle.
🚨🚨🚨 Call for papers alert 🚨🚨🚨 The "Differentiable Systems and Scientific Machine Learning" workshop at the 1st EurIPS conference is now accepting submissions! Let's explore the intersection of differentiable programming and SciML together. https://t.co/ligvk029Jl
0
1
13
🚨🚨🚨 Call for papers alert 🚨🚨🚨 The "Differentiable Systems and Scientific Machine Learning" workshop at the 1st EurIPS conference is now accepting submissions! Let's explore the intersection of differentiable programming and SciML together. https://t.co/ligvk029Jl
1
4
9
Journal of Open Source Software just published our "Tesseract Core: Universal, autodiff-native software components for Simulation Intelligence", @dionhaefner @AlexLavin_C137 🎉 https://t.co/OiZJ6S88bj
joss.theoj.org
Häfner et al., (2025). Tesseract Core: Universal, autodiff-native software components for Simulation Intelligence. Journal of Open Source Software, 10(111), 8385, https://doi.org/10.21105/joss.08385
0
3
6
Amazing event for people at ICML. Sad I cannot attend.
The JAX team is hosting a dinner / networking event during ICML on Thursday. Join us for an evening of food, drinks, and discussion of all things JAX. @SingularMattrix and other JAX team member will be attending. Please register early as capacity is limited. RSVP:
0
0
10
This week I started as a Research Scientist Intern at Meta Reality Labs in wonderful Zurich, Switzerland🇨🇭. Let me know if you are around and want to chat about neural emulators, differentiable physics, numerical solvers, etc.
10
5
283
We will use APEBench to train, test and benchmark it in an advection scenario against a feedforward ConvNet. https://t.co/AUfQNv7GTi
arxiv.org
We introduce the Autoregressive PDE Emulator Benchmark (APEBench), a comprehensive benchmark suite to evaluate autoregressive neural emulators for solving partial differential equations. APEBench...
0
0
5
Check Out my latest video on implementing an attention-based neural operator/emulator (i.e. a Transformer) in JAX:
3
12
100
I wanted to highlight PBDL's brand-new sections on diffusion models with code and derivations! Great work by Benjamin Holzschuh, with neat Jupyter notebooks 👍 All the way from normalizing flow basics over score matching to denoising & flow matching. E.g. https://t.co/TSDjuRPGwa
1
9
38