Skip to main content
Search

Federico Torrielli

1 related articles

Subscribe to updates
  1. arXiv

    Ranking positions by chat structure lets auditors explain 5% of tokens and keep nearly all threat-detection success

    Across about 4.7 million natural language autoencoder (NLA) explanations on four models and four auditing datasets, the study compares thirteen position-scoring signals computable in a single forward pass against a ranker trained only on chat structure, finding that chat structure beats the best individual signal in twelve of fourteen model-dataset cells, that explaining just 5% of positions retains nearly all of the success rate from explaining every position on three of four datasets, and that pretrained verbalizers recover words a fine-tuned model learned to conceal without additional verbalizer training.

You’re all caught up