AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

SafeSAE-VLA: Interpreting OpenVLA Progress Dynamics with Sparse Feature Analysis

Socrates Osorio, Joy Yang

ECCV 2026 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 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
use SAE top-20 features instead of full 16384-d SAE (ablation)SAE progress probe · LIBERO progress quartilesAUROC: 0.918 -> 0.894Hurt
SAE features versus motion-only controls (ablation)SAE progress probe · LIBERO progress quartilesAUROC: 0.572-0.711 -> 0.918Improved
dense raw-activation LR instead of SAE LR (ablation)progress probe · pooled encodingAUROC: 0.947 -> 0.975Improved
choose top-ranked 20 SAE features instead of random 20 (ablation)SAE LR · pooled encodingAUROC: 0.783 -> 0.926Improved
layer 20 versus layers 16 and 24 (ablation)SAE monitor · LIBERO progressAUROC: 0.884 / 0.896 / 0.871No clear effect
tertile or median split instead of quartile split (ablation)full SAE · LIBERO progressAUROC: 0.948 tertile, 0.929 medianNo clear effect
top-20 SAE instead of full dictionary under task-family shift (ablation)SAE readout · held-out tasksAUROC: 0.537 -> 0.658Improved
set top-20 features to high-progress class means (ablation)OpenVLA · raw-activation progress readoutreadout shift: +0.763 vs +0.078 randomImproved
set 13 active features to class mean (ablation)OpenVLA · LIBERO goal-task-1any-violation incidence: 1.00 -> 0.50Improved
progress-tuned top-20 used to predict success (ablation)Top-20 SAE · LIBERO success auditAUROC: 0.968 (full SAE) -> 0.471 (top-20)Hurt

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