I find it fascinating how they were able to keep the reconstruction error function incredibly simple, literally its success in round-tripping the activation layer, while making it interpretable... simply by choosing a good data-driven initialization state, and (effectively) training slowly.

I guess "initialization is all you need!"

From the paper https://transformer-circuits.pub/2026/nla/index.html :

> We find that simply initializing the AV and AR as copies of M leads to unstable training: the AV in particular, having never encountered a layer-l activation as a token embedding, outputs nonsensical explanations. We therefore initialize the AV and AR with supervised fine-tuning on a text-summarization proxy task. Specifically, we compute layer-l activations from the final token of randomly truncated pretraining-like text snippets, and use Claude Opus 4.5 to generate summaries s of the text up to that token (see the Appendix for details of this procedure). We then fine-tune the AV and AR on (h_l,s) and (s,h_l) pairs respectively. This warm-start typically yields an FVE of around 0.3-0.4. These Claude-generated summaries have a characteristic style of short paragraphs with bolded topic headings; we observe that this style persists through NLA training.

And from the appendix:

> We generate warm-start data for the AV and AR by prompting Claude Opus 4.5 to produce summaries of contexts, using the prompt below. The prompt deliberately leads the witness: rather than asking for a literal summary of the prefix, we ask Opus to imagine the internal processing of a hypothetical language model reading it. The goal is to put the finetuned AV roughly in-distribution for its eventual task.