AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders

David Chanin, Adrià Garriga-Alonso

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 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
set SAE L0 to the true L0 instead of too low (1.8 vs 2) (ablation)BatchTopK SAE · Toy model (5 features)feature recovery: L0 1.8 -> L0 2Improved
use ground-truth SAE instead of trained low-L0 SAESAE at L0=5 · Toy model reconstructionMSE: 2.73 -> 4.88Hurt
start training from too-high L0 (20) then decay to correct L0 (11) (ablation)BatchTopK SAE · Toy model (50 features)feature recovery: L0 20 -> 11No clear effect
start training from too-low L0 (2) then raise to correct L0 (11) (ablation)BatchTopK SAE · Toy model (50 features)feature recovery: L0 2 -> 11Hurt
train SAE with L0 too high (18) instead of correct (11) (ablation)BatchTopK SAE · Toy model (50 features)feature recovery: L0 11 -> 18Hurt
train SAE with L0 too low (5) instead of correct (11) (ablation)BatchTopK SAE · Toy model (50 features)feature recovery: L0 11 -> 5Hurt
use JumpReLU instead of BatchTopK at high L0SAE · Gemma-2-2b layer 12k-sparse probing F1Improved
use L0 near 200 (c_dec elbow) instead of lower L0BatchTopK SAE · Gemma-2-2b layer 5 and Llama-3.2-1b layer 7k-sparse probing F1: elbow around L0 200Improved

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