The best Llm Interpretability newsletters

A researcher shares a paper proposing persistent homology as a multi-scale framework for characterising the topology and geometry of LLM latent spaces, with findings on adversarial influence. The newsletter is a brief announcement of that work, promising a longer post later.