Since people have been asking - the #blackboxNLP workshop will return this year, to be held with #emnlp2025. This workshop is all about interpreting and analyzing NLP models (and yes, this includes LLMs). More details soon, follow BlackboxNLP

Our paper "Position-Aware Circuit Discovery" got accepted to ACL! 🎉 Huge thanks to my collaborators🙏 Hadas Orgad @ ICML David Bau Aaron Mueller Yonatan Belinkov See you in Vienna! 🇦🇹 #ACL2025 ACL 2025

Great work on unlearning in LLMs—very neat method and an impressive contribution.

Tried steering with SAEs and found that not all features behave as expected? Check out our new preprint - "SAEs Are Good for Steering - If You Select the Right Features" 🧵

Congrats Hadas Orgad and Alex on your beautiful wedding! Wishing you a lifetime of joy, love, and highly-cited collaborations 💍✨

Preferences drive modern LLM research and development: from model alignment to evaluation. But how well do we understand them? Excited to share our new preprint: Multi-domain Explainability of Preferences arxiv.org/abs/2505.20088 Roi Reichart Liat 🧵👇 1/11

VLMs perform better when answering questions about text than when answering the same questions about images - but why? and how can we fix it? We investigate this gap from a mechanistic interpretability perspective, and use our findings to close a third of it! 🧵

After discussing the Llada paper today, we were wondering: is this just masked language modeling Bert style? Main differences seem to be: (a) training with varying masking budgets; (b) inference with gradual unmasking determined by confidence. arxiv.org/abs/2502.09992

Now accepted to #COLM2025! We formally define hidden knowledge in LLMs and show its existence in a controlled study. We even show that a model can know the answer yet fail to generate it in 1,000 attempts 😵 Looking forward to presenting and discussing our work in person.

🚨New paper alert🚨 🧠 Instruction-tuned LLMs show amplified cognitive biases — but are these new behaviors, or pretraining ghosts resurfacing? Excited to share our new paper, accepted to CoLM 2025🎉! See thread below 👇 #BiasInAI #LLMs #MachineLearning #NLProc

BlackboxNLP is the workshop on interpreting and analyzing NLP models (including LLMs, VLMs, etc). We accept full papers and extended abstracts. The workshop is highly attended; great exposure for your finished work or feedback on work in progress. #emnlp2025 at Sujhou, China!

Announcing a deep net interpretability talk series! Every week you will find new talks on recent research in the science of neural networks. The first few are posted: @jack_merulllo_, Roy Rinberg, and me. At the NDIF Youtube Channel: youtube.com/@NDIFTeam.

Excited to share our recent work: DeLeaker! We're tackling semantic leakage in T2I models.

At the #Neurips2025 mechanistic interpretability workshop I gave a brief talk about Venetian glassmaking, since I think we face a similar moment in AI research today. Here is a blog post summarizing the talk: davidbau.com/archives/2025/…

Interested in changes in perception of the term NLP vs LLMs. Which statement do you agree with? - NLP and LLMs are just different things these days - LLMs are a subset of NLP - NLP is deprecated due to LLMs