Sam Marks comments on Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations

[Sam Marks]

The full report has some evaluations for steganography, measuring how much reconstruction falls when paraphrasing explanations in different ways.