bennokrojer.bsky.social
AI PhDing at Mila/McGill. Happily residing in Montreal 🥯❄️ Academic stuff: language grounding, vision+language, interp, rigorous & creative evals, cogsci Other: many sports, urban explorations, puzzles/quizzes bennokrojer.com
👤 Joined on 24 April 2023
Followers
Following
Total Posts
@bennokrojer.bsky.social@bennokrojer.bsky.social → @mcgill-nlp.bsky.social@mcgill-nlp.bsky.social → @bennokrojer.bsky.social[New preprint] How do images in VLMs align with words? In this work, we found a set of attention heads responsible for OCR. But to our surprise, these heads were actually able to verbalize much more than just text. So, we used them to create a simple logit lens for image tokens!
Show PostThanks to co-authors @bennokrojer.bsky.social, Sarah Wang, Henry Abrahamsen, @byron.bsky.social, and @davidbau.bsky.social for helping get this paper done. :) Website: ocr.baulab.info/ arXiv: arxiv.org/abs/2609.18823 Code: github.com/sfeucht/ocr
Show Post@bennokrojer.bsky.social is here too btw ^^
Show PostJust uploaded a new plant blog post on my website 🥰🪴🌸
Show Post