Internalizing RoPE embeddings. If, like me, you slept under the rock since the introduction of RoPE in 2021 this document provides the most-brief late 2025 update on the latest practical tech. Jump start your paper reading here!

Over a 1000 publications and posters has been presented at NIPS2017. Here I would like to mention just a few most fascinating works. I believe that these publications have a fresh view on deep learning. They are promising a great deal of progress in the field.

NIPS2017 is over. It has been 6 intense days full of philosophy, research, science, inspiration and idea exchange. It is time to organize the knowledge now. Give it some structure and document some outtakes while they are still fresh.