AI Paper Atlas

Research map / Probabilistic and scientific modeling / Causal discovery

Readout Representation: Redefining Neural Codes by Input Recovery

Shunsuke Onoo, Yoshihiro Nagano, Yukiyasu Kamitani

ICLR 2026 · Probabilistic and scientific modeling › Causal discovery · approach: Causal representation identifiability · 1 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
perturb VGG19 lower-layer features to 0.7 correlation distance before inversion (ablation)feature inversion readout · VGG19 lower layerspixel-space correlation distance of reconstruction: 0.7 feature perturbation still within 0.1 pixel distanceNo clear effect
use features from higher layers instead of lower-to-middle layers (ablation)feature inversion from perturbed features · VGG19, DINOv2, CLIPrecovery under perturbationHurt
use GPT2 (and some OPT variants) instead of BERT/OPT-350m (ablation)text feature inversion · C4 textstoken error rate under perturbationHurt
compare correctly classified vs misclassified imagesrepresentation size · VGG19, ImageNet val (8 vs 8 images)representation sizeImproved
use random uniform noise images instead of natural imagesrepresentation size · VGG19representation size: zero across all layersHurt
increase feature dimensionality of layerrepresentation size · vision models across layersrepresentation sizeMixed
remove DIP and invert in pixel space (ablation)representation size · VGG19representation sizeHurt

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