Papers I wish I could get you to read

2 issues • one-page explainers

1. 
   Decompiling Transformers

   #002Discovering Interpretable Algorithms by Decompiling Transformers to RASP • Huang, Bakalova et al. | ICML 2026
   ···
2. 
   Tracr

   #001Compiled Transformers as a Laboratory for Interpretability • Lindner et al. (Google DeepMind) | NeurIPS 2023
   ···