AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Depth-Progressive Monotonic Learning without Global Backpropagation

Chenhao Ye, Rongguang Ye, Yuchao Zhang, Ming Tang

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse

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
replace global backprop with DMoL local belief objectivesSimpleCNN · CIFAR-100accuracy (%): 51.81 -> 56.12Improved
replace global backprop with DMoLSimpleCNN · Tiny-ImageNetaccuracy (%): 34.51 -> 41.37Improved
replace global backprop with DMoLSimpleCNN / ViT-Tiny · CIFAR-10 and ImageNetaccuracy (%): 84.32 -> 84.42 (CIFAR-10); 63.68 -> 63.74 (ImageNet)No clear effect
replace BP with DMoL in 8-layer GCN to reduce over-smoothing8-layer GCN · Coraaccuracy (%): 43.86 -> 81.40Improved
replace BP with DMoL in 8-layer GCN8-layer GCN · Actoraccuracy (%): 25.46 -> 27.78Improved
replace BP with DMoL for diffusion transformer trainingDiT · CIFAR-100 label-to-imagefinal training MSE: 0.060 -> 0.034Improved
use standard ResNet-18 with DMoL instead of BPResNet-18 · CIFAR-10accuracy (%): 95.68 -> 93.32Hurt
adapt ResNet-18 to DMoL (ResNet-DMoL-RA variant) (ablation)DMoL on ResNet-18 · CIFAR-100accuracy (%): 73.02 -> 79.16Improved
increase network depth from 4 to 20 modules (ablation)DMoL vs BP · CIFAR-100accuracy (%)Mixed
remove then restore one-third of intermediate modules during training (ablation)BP vs DMoL · structural fragility testtest accuracyMixed
add parallel workers for module updatesDMoL vs BP · CIFAR-100throughput speedupImproved

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