In general, all of Anthropic’s work is super interesting to read.

In general, all of Anthropic’s work is super interesting to read. And today there’s something new.
On interpretability
a new series just came out. (https://www.anthropic.com/news/mapping-mind-language-model)
The previous series (https://transformer-circuits.pub/2021/framework/index.html#mathematical-framework)
I consider it one of the best works out there on understanding how a transformer works.