AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Causal Interpretation of Neural Network Computations with Contribution Decomposition (CODEC)

Joshua Brendan Melander, Zaki Alaoui, Shenghua Liu, Surya Ganguli, Stephen Baccus

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Mechanistic circuit interpretability

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
decompose contributions instead of activations with SAE (ablation)CODEC on ResNet-50 · ImageNet val, class correlation of modesmode-class correlationImproved
use contribution modes instead of individual channels (ablation)CODEC on ResNet-50 · ImageNet val, class correlationcorrelation with classImproved
select channels by contribution modes rather than activation modes for ablation/preservation (ablation)ResNet-50 network control · ImageNet, target vs off-target classablation specificityImproved
analyze contributions instead of activationsResNet-50 layers · ImageNet valHoyer sparsityImproved
analyze contributions instead of activationsResNet-50 class-averaged channel weightings · ImageNet valdimensionality (components for 95% variance)Improved
set dictionary size to equal or smaller than channel count (ablation)CODEC SAE · ResNet-50 layers 3,7,13,15SAE R^2 / stabilityHurt
raise SAE threshold from 0.5 to 0.9 (ablation)CODEC SAE · ResNet-50SAE R^2Hurt
vary L1 regularization, random seed, non-negativity constraint (ablation)CODEC SAE · ResNet-50SAE reconstruction R^2No clear effect
shrink SAE encoder MLP hidden dimension (ablation)CODEC SAE · ResNet-50reconstructionHurt
apply CODEC contribution-SAE instead of activation-SAE on ViT-B (ablation)ViT-B token, MLP and attention layers · ImageNet ablationablation specificityImproved
use contrastive target instead of top-1 logit (ablation)CODEC · ResNet-50 four layerssparsity: R^2 above 0.85No clear effect
group ViT attention dimensions into learned heads (ablation)ViT-B attention · ImageNetHoyer sparsity vs random groupingNo 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