FWIW, this line of research seems to have largely been shown to be ineffective for model steering in practice! arxiv.org/abs/2501.17148 from @aryaman.io is my reference but several others have shown similar results!

Show Post

Unlike in Indo-Aryan languages where all reflexes of Sanskrit rā́trī (& rātram) seem to be a frozen form spread from the Gangetic Plains (e.g., Gawri rā̀t instead of **rā́ɬ, Shina rā́ti instead of **rā́c̣(i), Sindhi rātĭ instead of **rāṭĭ, Romani rat instead of **ratr-, Kalasha rat instead of **ratr), ...

Show Post

"Mission: Impossible" was featured in Quanta Magazine! Big thank you to @benbenbrubaker.bsky.social for the wonderful article covering our work on impossible languages. Ben was so thoughtful and thorough in all our conversations, and it really shows in his writing!

Show Post

And a big thank you to everyone who came to the talk itself. The discussion period after was really rich and wide-ranging.

Show Post

I thank lots of people at the very end for their role in shaping this work. A special shout-out to @aryaman.io for creating CausalGym, which made it very easy for me to conduct all the intervention-based analysis in the talk: github.com/aryamanarora...

Show Post