The Information Machine

New Formal Methods for Reading Model Internals From Weights

Synthesis history

2 versions, newest first.

  1. Version 2 2026-05-16 04:40 UTC · 5 items

    No substantive new items arrived this pass. The three new entries (items 7232, 1947, 1958) are generic Wikipedia background articles on neural network history, large language models, and machine learning — they carry no…

  2. Version 1 2026-05-12 20:14 UTC · 2 items

    Two research groups are independently developing methods to extract behavioral properties of neural networks directly from their weights — without running the model on inputs. Lucius Bushnaq (affiliated with the parame…