🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

We're excited to have Sébastien Lachapelle presenting his work on additive decoding and extrapolation at the CARE reading group tomorrow! bit.ly/3x3kuRu

📢 I'm recruiting a postdoc at UMD 🗓️ Starting in 2024 🖥️🙋 Topics related to human-AI interaction 🔜 Apply by May 31, but earlier if possible! 🤑 $75-$80k+benefits, unencumbered (ie not "big project") funding Link to full post and form for applications in thread. >

This week at CARE (Thursday 11am EST) we have Atticus Geiger presenting his fantastic line of work (with Christopher Potts, Thomas Icard, noahdgoodman, and others) on finding causal abstractions of large language models. Details here: portal.valencelabs.com/events/post/un…

CARE is back this week, and we'll have the pleasure of hearing about Atticus' super cool line of work on making causal abstraction practical in service of interpreting large models.

👀 Getting to learn about causal abstraction from the OG behind causal abstraction?? Don't sleep on this!

📣 Postdoctoral fellowships, up to 24 months to conduct cutting-edge research in Europe. If you would like the Rational Intelligence Lab life to host you, reach out to me directly. 🤝 🗓️ Deadline: 11 September 2024 🙏 RT please 🥺 …sklodowska-curie-actions.ec.europa.eu/news/msca-open…

Join us at CARE today at 11am EST to hear Anish Agarwal talk about a staple of causal inference, synthetic control and intervention. This marks our foray into extrapolation.

This week at CARE, as a treat after the NeurIPS ddl, we have Sorawit (James) Saengkyongam about learning representations for extrapolation! Details here: bit.ly/3UIcM7y

Congratulations to Mila members Aishwarya Agrawal Golnoosh Farnadi @NeurIPS2024 Dhanya Sridhar Siva Reddy for pushing the boundaries of AI research on generative models, auditing tools, algorithmic fairness and cultural diversity in natural language processing!

Very excited about this work and other cool research that's brewing around pre-trained autoregressive models and ICL 😎

🇨🇦Yesterday kicked off our annual AICan meeting, bringing together Canada CIFAR AI Chairs from Canada’s three National AI Institutes, Amii, Mila - Institut québécois d'IA and Vector Institute, for 3 days of discussion, collaboration and learning.

📢 Nidhi and I are looking for a postdoc to work on strategic classification and causal modeling. Nidhi is at ICML so reach out to her if you're there and want to chat about this opportunity. For more details and to apply: nidhihegde.github.io/postdoc2024/

Friends! I am looking for PhD and Masters students to join my group for Fall 2025, Please share with students you may know who are interested in the intersection of language, CogSci, and AI! first deadline is Dec 1st, see my page for more info: evaportelance.github.io/prosstudents.h…

📣Call for Blog Posts at #ICLR2025 ICLR 2026 Following the success of the past iterations, we are opening the Call for Blog Posts 2025! iclr-blogposts.github.io/2025/call Please retweet!

I am hiring Ph.D. and/or M.Sc. students to work at the intersection of game theory, optimization, and Machine Learning in Fall 2025. If you are interested in working in my group, apply via mila.quebec/en/prospective… More details on the topics in the 🧵👇 1/n

Join Mila for PhD/MSc! 🇨🇦 Apply in link below I will consider all candidates, with extra attention to unusual backgrounds, underrepresented groups, regions under war, occupation, instability (Ukraine, Palestine, Africa, … ) Optionally, if your background is unusual DM me here

I'll be looking for 1-2 students this year (co-supervision included). See my website for details about my research and stay tuned for a page coming soon about my evaluation process. Apply to Mila by Dec 1!

Awesome new work from our ICL working group (give these folks a follow)! The takeaway is that next token prediction leads to in-context learners that satisfy Occam's razor, i.e., simple prediction functions.

Awesome work led by my student Shruti Joshi on steering generative models without the need for carefully collected contrastive pairs! Tl;dr: We link identifiability theory to sparse autoencoding, allowing us to recover steering vectors from multi-concept shift data.