Hannah Pinson (@pinson_h) 's Twitter Profile
Hannah Pinson

@pinson_h

Assistant Prof. of AI @ Eindhoven University of Technology, guest Prof. at VUB. Previously VUB & MIT. Interested in fundamental insights about deep learning.

ID: 3972230793

linkhttps://hannahpinson.be/ calendar_today16-10-2015 13:26:21

313 Tweet

770 Takipçi

738 Takip Edilen

Data and AI cluster @ TU/e (@data_ai_tue) 's Twitter Profile Photo

📢Job alert! Interested in robustness and reliability of AI? Check out the PhD candidate vacancy in our Uncertainty in AI group at TU/e: jobs.tue.nl/en/vacancy/phd… #PhD #vacancy #AI

Hannah Pinson (@pinson_h) 's Twitter Profile Photo

Today I woke up to this view on the gulf of Trieste 😍, a great start to the #YIHD2024 conference on Machine Learning, High-Dimensional Statistics and Inference! My talk was recorded and I hope to share it later😊 Thanks Sebastian Goldt and other organizers for the invite.

Today I woke up to this view on the gulf of Trieste 😍, a great start to the #YIHD2024 conference on Machine Learning, High-Dimensional Statistics and Inference! 

My talk was recorded and I hope to share it later😊 Thanks <a href="/sebastiangoldt/">Sebastian Goldt</a> and other organizers for the invite.
Hannah Pinson (@pinson_h) 's Twitter Profile Photo

💡Which information about the dataset does a simple CNN, trained with GD, store in its parameters? Only the most important frequencies, through a winner-takes-all process! Judging from my body language, I was trying hard to get the message across 😅 youtube.com/watch?v=YJPmwk…

💡Which information about the dataset does a simple CNN, trained with GD, store in its parameters? Only  the most important frequencies, through a winner-takes-all process!

Judging from my body language, I was trying hard to get the message across 😅 youtube.com/watch?v=YJPmwk…
Andrew Saxe (@saxelab) 's Twitter Profile Photo

📢 Job alert! 📢 Postdoctoral fellowship in deep learning theory, with a focus on learning dynamics in modern architectures like transformers. Join our wonderful community Gatsby Computational Neuroscience Unit SWC More info: ucl.ac.uk/work-at-ucl/se… Applications reviewed from: June 30 Please RT!

Mykola Pechenizkiy (@pechenizkiy) 's Twitter Profile Photo

We have exciting postdoc positions on LLMs in Software Engineering in collaboration with ASML: jobs.tue.nl/en/vacancy/two… Please share with suitable candidates.

Andres Algaba (@andresalgaba1) 's Twitter Profile Photo

🧵Our new study (arxiv.org/abs/2405.15739) reveals how LLMs like GPT-4 exhibit human-like citation patterns but with increased citation bias. Let’s dive into the findings. A thread🧵. Data Analytics Lab Carmen Mazijn Vincent Holst Floriano Tori Vincent Ginis  Sylvia Wenmackers

Stefano Sarao Mannelli (@stefsmlab) 's Twitter Profile Photo

Very excited to announce openings for my first 2 lab members Chalmers University of Technology 1️⃣ PhD in theory of ML bias [chalmers.se/om-chalmers/ar…] 2️⃣ Postdoc in ML theory [chalmers.se/om-chalmers/ar…] Beautiful campus, a liveable city near an archipelago 🛶 Work closely with me to shape the lab! 🌱

Data and AI cluster @ TU/e (@data_ai_tue) 's Twitter Profile Photo

Jiarong is currently presenting his (spotlight! 🎉) work called MALIBO at #ICML2024, with co-authors Stefan Falkner, Felix Berkenkamp, and Joaquin Vanschoren. 🥳Our group members are presenting 13 papers at ICML, see dai.win.tue.nl/newsitems/28/

Jiarong is currently presenting his (spotlight! 🎉) work called MALIBO at #ICML2024, with co-authors Stefan Falkner, Felix Berkenkamp, and <a href="/joavanschoren/">Joaquin Vanschoren</a>. 

🥳Our group members are presenting 13 papers at ICML, see dai.win.tue.nl/newsitems/28/
Hannah Pinson (@pinson_h) 's Twitter Profile Photo

Tomorrow Friday I'll present "Exploring the development of complexity over depth and time in deep neural networks" as a poster at the #ICML2024 Workshop on High Dimensional Learning Dynamics (HiLD) 🥳 See you in Straus 2 (across Hall B towards Hall C) 10-11 am and 3.30-4 pm!

Tomorrow Friday I'll present "Exploring the development of complexity over depth and time in deep neural networks" as a poster at the #ICML2024 Workshop on High Dimensional Learning Dynamics (HiLD) 🥳

See you in Straus 2 (across Hall B towards Hall C) 10-11 am and 3.30-4 pm!
Hannah Pinson (@pinson_h) 's Twitter Profile Photo

I'm looking for research about the evolution of LLM/transformer "hallucinations" during training: how does the nature of hallucinations change while the network learns? All suggestions are welcome!