update: added a few readings and direction, fixed typos and sentence structures I became enormously interested in this aspect of LLMs recently due to the simple question on how models might behave when they are adapted to new data, task, and knowledge. This question came up after watching my colleagues talk about Test Time Adaptation on LLMs, and I immediately wondered how it affects alignment measures that are in place in the base model. However, after further review, I realized that TTA makes little to no sense (at least in the context of what we might imagine an agent that continuously adapts would do) and seems to be lazily motivated (meant to just extend TTA into the LLM domain).
A month ago, Jane Street came to KAIST to present their company and what their employees did, which was quite fascinating. They also hosted an estimathon, and divided us into random groups to compete. It was quite fun! I solved some problems and I think our team did alright, but unfortunately we didn’t get the overall win. At the end of the session, they gave us all some stickers and a t-shirt with a nice little problem on it