It's on arXiv! (written 10 June)
Last year, I spent half a year in Zurich at ETH, and in the second class of Ryan Cotterell’s excellent Large Language Models course I walked up to him to ask if he was looking for research assistants. It has since been a year with several ups and downs, but today, we (meaning myself, Josef Valvoda, Tianyu Liu, Anej Svete and Ryan Cotterell) put up a paper on arXiv, titled somewhat dramatically “A Fundamental Tradeoff in Aligned Language Models and its Relationship to Sampling Adaptors”. You can find it here.
EMNLP (written 14 June)
We’ve submitted it for EMNLP 2024 :)