iiinova — where your digital self keeps evolving

Reposted by Nils Trost

Lena Kutscher @lenakutscher.bsky.social · 10/02/2026

Excited to share our preprint on our new multi-omic atlas of human hindbrain development. Led by postdoc Piyush Joshi, in collaboration with @kaessmannlab.bsky.social and Pfister labs, our atlas represents the first comprehensive view of human hindbrain development. www.biorxiv.org/content/10.6...

Brain with puzzle overlay to show that our study provides missing pieces of the puzzle of human brain development by delivering the most comprehensive picture of hindbrain development to date. We have strived to go beyond just another multi-omics atlas to gain deep insights by:
1. Meticulously annotating cell clusters
2. Extracting regulatory programs in terms of coordinated gene sets and accessible regulatory elements
3. Using deep learning to identify regulatory syntax
4. Resolving context-specific TF activity

23315

Reposted by Nils Trost

Universität Heidelberg @uniheidelberg.bsky.social · 14/11/2025

Computerlinguistin und Geoinformatiker mit „KlarText-Preis“ ausgezeichnet – Hervorragende Wissenschaftskommunikation: Letitia Parcalabescu und Steffen Knoblauch präsentieren Ergebnisse ihrer Doktorarbeiten in allgemeinverständlicher Form www.uni-heidelberg.de/de/newsroom/... #wisskomm

Die Computerlinguistin Dr. Letitia Parcalabescu und der Geoinformatiker Dr. Steffen Knoblauch

052

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 02/11/2025

LLMs can memorize even a phone number seen once in training.🔒 Google’s VaultGemma fixes that, being the first open-weight LLM trained from scratch with differential privacy, so rare secrets leave no trace. ☕ new video explaining Differential Privacy through VaultGemma 👇 🎥 youtu.be/UwX5zzjwb_g

youtu.be

What's up with Google's new VaultGemma model? – Differential Privacy explained

YouTube video by AI Coffee Break with Letitia

062

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 19/10/2025

We explain diffusion models and flow-matching models side by side. Flow-Matching models are the new generation of AI image generators that are quickly replacing diffusion models. They take everything diffusion did well, but make it faster, smoother, and deterministic. 🎥 youtu.be/firXjwZ_6KI

youtu.be

Diffusion Models and Flow-Matching explained side by side

YouTube video by AI Coffee Break with Letitia

021

Reposted by Nils Trost

Kaessmann Lab @kaessmannlab.bsky.social · 16/10/2025

Did transposable elements shape brain evolution — and if so, which ones, and in which cell states and lineages? Led by @tyamadat.bsky.social, we explored this question in cerebellum development using sequence-based deep learning models! www.biorxiv.org/content/10.1...

58135

Reposted by Nils Trost

Kaessmann Lab @kaessmannlab.bsky.social · 15/10/2025

We are thrilled to share our new preprint entitled “The origin and molecular evolution of the mammalian liver cell architecture” www.biorxiv.org/content/10.1...

35419

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 21/09/2025

Ever wondered how Energy-Based Models (EBMs) work and how they differ from normal neural networks? ☕️ We go over EBMs and then dive into the Energy-Based Transformers paper to make LLMs that refine guesses, self-verify, and could adapt compute to problem difficulty.

3226

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 14/09/2025

The world’s largest NLP conference with almost 2,000 papers presented, ACL 2025 just took place in Vienna! 🎓✨ Here is a quick snapshot of the event via a short interview with one of the authors whose work caught my attention. 🎥 Watch: youtu.be/GBISWggsQOA

032

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 03/08/2025

How do LLMs pick the next word? They don’t choose words directly: they only output word probabilities. 📊 Greedy decoding, top-k, top-p, min-p are methods that turn these probabilities into actual text.

132

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 27/07/2025

Excited to be at ACL 2025 in Vienna this week 🇦🇹 #ACL2025 I’m always up for a chat about reasoning models, NLE faithfulness, synthetic data generation, or the joys and challenges of explaining AI on YouTube. If you're around, let’s connect!

281

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 07/07/2025

📅 Looking forward to the discussion and to learning from fellow panelists and participants. If you're around Heidelberg, join us! www.marsilius-kolleg.uni-heidelberg.de/de/dancing-w...

marsilius-kolleg.uni-heidelberg.de

Dancing with Right & Wrong? - Marsilius-Kolleg

An interdisciplinary symposium on knowing, doing and not being sure

011

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 07/07/2025

🤖 Can we trust AI in science? I'm excited to be speaking at the final event of the Young Marsilius Fellows 2025, themed "Dancing with Right & Wrong?" – a title that feels increasingly relevant these days. I'll be joining a panel on "(How) can we trust AI in science?" to discuss questions like:

121

Nils Trost @trostnils.bsky.social · 20/06/2025

I'm very excited to finally share the main work of my PhD! We explored the evolutionary dynamics of gene regulation and expression during gonad development in primates. We cover among others: X chromosome dynamics (incl. in a developing XXY testis), gene regulatory networks and cell type evolution.

1145

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 31/05/2025

Excited to share that I’ll be joining the Summer School “AI and Human Values” this September at the Marsilius-Kolleg of Heidelberg University as a speaker. I'll be giving an introduction to how large language models actually work—before the summer school dives deeper into broader implications.

132

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 18/05/2025

Long videos are a nightmare for language models—too many tokens, slow inference. ☠️ We explain STORM ⛈️, a new architecture that improves long video LLMs using Mamba layers and token compression. Reaches better accuracy than GPT-4o on benchmarks and up to 8× more efficiency. 📺 youtu.be/uMk3VN4S8TQ

youtu.be

Token-Efficient Long Video Understanding for Multimodal LLMs | Paper explained

YouTube video by AI Coffee Break with Letitia

052

Reposted by Nils Trost

Kaessmann Lab @kaessmannlab.bsky.social · 07/04/2025

A great article highlighting the broader implications of the three parallel recent papers from our group and our wonderful colleagues in Spain and Belgium...

0287

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 23/03/2025

Just say “Wait…” – and your LLM gets smarter?! We explain how just 1,000 training examples + a tiny trick at inference time = o1-preview level reasoning. No RL, no massive data needed. 🎥 Watch now → youtu.be/XuH2QTAC5yI

youtu.be

s1: Simple test-time scaling: Just “wait…” + 1,000 training examples? | PAPER EXPLAINED

YouTube video by AI Coffee Break with Letitia

043

Reposted by Nils Trost

Kaessmann Lab @kaessmannlab.bsky.social · 16/03/2025

How does gene regulation shape brain evolution? Our new preprint dives into this question in the context of mammalian cerebellum development! rb.gy/dbcxjz Led by @ioansarr.bsky.social, @marisepp.bsky.social and @tyamadat.bsky.social, in collaboration with @steinaerts.bsky.social

419270

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 19/01/2025

An educational and a bit historical deep dive into LLM research. 💡Learn what breakthroughs since 2017 paved the way for AI like ChatGPT (it wasn't overnight). We go through: * Transformers * Prompting * Human Feedback, etc. and break it all down for you! 👇 📺 youtu.be/BprirYymXrg

youtu.be

LLMs Explained: A Deep Dive into Transformers, Prompts, and Human Feedback

YouTube video by AI Coffee Break with Letitia

031

Reposted by Nils Trost

aicoffeebreak.bsky.social @aicoffeebreak.bsky.social · 08/12/2024

New video about: REPA (Representation Alignment), a clever trick to align diffusion transformers’ representations with pretrained transformers like DINOv2. It accelerates training and improves the diff. model’s ability to do things other than image generation (like image classification).

youtu.be

REPA Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You ...

YouTube video by AI Coffee Break with Letitia

131