AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Discovering Interpretable Algorithms by Decompiling Transformers to RASP

Xinting Huang, Aleksandra Bakalova, Satwik Bhattamishra, William Merrill, Michael Hahn

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and learnability · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply causal pruning to the D-RASP translationD-RASP program, 4-layer 4-head transformer · Most Frequentprogram lines: 16,201,616 -> 3Improved
apply causal pruning to the D-RASP translationD-RASP program, 1-layer 4-head transformer · Most Frequentprogram lines: 56 -> 3Improved
apply decompilation to length-generalizing modelsDecompilation pipeline · Formal languagesdecompilation success: 9 of 14 languagesMixed
apply decompilation to non-length-generalizing modelsDecompilation pipeline · Formal languagesdecompilation success: 0 of 3 languagesHurt
apply decompilation to non-length-generalizing modelsDecompilation pipeline · Algorithmic tasksdecompilation success: 0 of 3 tasksHurt
apply decompilation to length-generalizing modelsDecompilation pipeline · Algorithmic tasksdecompilation success: 8 tasksImproved
apply causal pruning to non-length-generalizing modelsCausal pruning · Algorithmic tasksprogram size reductionNo clear effect
linearize layer norm (ablation)Transformer · D3 Dyck languagedecompilation successHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy