AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Automated Interpretability Metrics Do Not Distinguish Trained and Random Transformers

Thomas Heap, Tim Lawson, Lucy Farnik, Laurence Aitchison

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 41 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
use trained instead of randomly initialized Pythia weights (ablation)TopK SAE (R=64, k=32) on Pythia residual stream · Pythia 70M-6.9B, fuzzing AUROCauto-interpretability AUROCNo clear effect
replace consistent token embeddings with i.i.d. Gaussian noise (control) (ablation)SAE on trained Pythia · Pythia, fuzzing AUROCauto-interpretability AUROCHurt
use trained instead of randomized weights (ablation)SAE on Pythia · Pythia, token-distribution entropy of latentsentropy over token IDsImproved
increase model size (ablation)SAE on randomized Pythia · Pythia 70m vs 6.9b, fuzzing AUROCauto-interpretability AUROC gap trained vs randomMixed
train SAEs on 1B tokens instead of 100M (ablation)SAE on Pythia-70m/410m · Pythia, SAE metricsevaluation metricsNo clear effect
vary expansion factor 16-128 and sparsity 16-32 (ablation)SAE on Pythia-160m · Pythia-160mSAE evaluation metricsNo clear effect

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