This is such a cool paper! “No computation without abstraction”

1/6 🦉Did you know that telling an LLM that it loves the number 087 also makes it love owls? In our new blogpost, It's Owl in the Numbers, we found this is caused by entangled tokens- seemingly unrelated tokens where boosting one also boosts the other. owls.baulab.info/

A short 📹 explainer video on how LLMs can overthink in humanlike ways 😲! had a blast presenting this at #icml2025 🥳

maybe I will live tweet the actionable interp workshop panel

We see so many work this week about "emergent misalignment", but how is it fundamentally different from LLM jailbreaking research? I wrote a short blog post about it: yongzx.substack.com/p/emerge…

【#ICML2025 Poster】 [1/7] Many works develop intricate “jailbreaks” that elicit harmful outputs from LLMs. But can more common user-LLM interactions cause the same? We show yes! Paper: arxiv.org/abs/2502.04322 Coauthors: @yiksiux, @YuxinXiao6, @MarzyehGhassemi

Lots of progress in mech interp (MI) lately! But how can we measure when new mech interp methods yield real improvements over prior work? We propose 😎 𝗠𝗜𝗕: a Mechanistic Interpretability Benchmark!

Thank you for featuring our work!!

I’m grateful to have been part of this collaboration on LLMs for health with the amazing team at MIT. Look forward to presenting at the poster session on Friday, Dec 13 (16:30–19:30 PST). Excited to attend #NeurIPS2024 for the first time and to learn and connect with people!