Back to papers
lesswrong8.0 / 10

Some observations about NLA explanations

loops

Abstract

I used the Gemma 3 12B activation verbalizer (maps activations to English) and reconstructor (maps English to activations) described in the Natural Language Autoencoders (NLA) paper to generate a bunch of explanations for 20k random tokens from a pretraining dataset (Common Pile derivative) and another 20k random tokens from a chat dataset . I also reconstructed all of the activations from the verbalizations so that I could see what kinds of tokens and explanations have high reconstruction error. I published pretraining token explanations and chat token explanations [1] and the code I used (written with Claude). Here are some thoughts I have about the explanations and some experiments I did with the data. Explanation text qualities Almost all of the Gemma explanations I read follow the same three part format, with each part is being 1-2 sentences. A sentence that describes the kind of document we're in, then a colon, then a description of what the document's about. For chat sessions this usually starts with the word "Structured" (80% for chat sessions vs 7% for non-chat). Two examples (non-chat vs chat): Literary narrative pattern: formal English countryside setting, with a medieva

Research area

post-trainingpretraining
Published
15 May 2026
Source
lesswrong
Org
Alignment Forum
View paper
Sign in to read and join the discussion.