AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Toward the Connection Between Activation Sparsity and Flat Minima

Ze Peng, Jian Zhang, Lei Qi, Yang Gao, Y Shi

TPAMI 2026 · Architectures and efficiency › RNNs and normalization · approach: Theory of depth and width · 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
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)ViT-B · ImageNet-1K pretrainingtraining sparsity: 0.104 -> 0.046Improved
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)ViT-B · ImageNet-1K pretrainingtesting sparsity: 0.087 -> 0.055Improved
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)ViT-B · ImageNet-1K pretrainingAcc@1: 77.35% -> 76.77%Hurt
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)SwinTransformer-B · ImageNet-1K pretrainingtesting sparsity: 0.057 -> 0.041Improved
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)SwinTransformer-B · ImageNet-1K pretrainingAcc@1: 79.75% -> 79.82%No clear effect
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)ViT-L · Places365 pretrainingtesting sparsity: 0.039 -> 0.018Improved
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)ViT-B · LAION-400M subsettest loss: 0.984 -> 1.067Hurt
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)T5-Base · C4 pretrainingtesting sparsity: 0.299 -> 0.180Improved
add sparsity-aware modifications (zeroth biases, lower-bounded LayerNorm, new activation)T5-Base · C4 pretrainingtest loss: 4.88 -> 4.78Improved
sparsity-aware LoRA finetuning instead of vanilla finetuningViT-Base · ImageNet-1K finetuningtesting sparsity / Acc@1: 0.122 -> 0.102 sparsity; 77.17% -> 74.46% Acc@1Mixed
sparsity-aware LoRA finetuning instead of vanilla finetuningT5-Base · C4 finetuningtesting sparsity: 0.304 -> 0.199Improved
add lower-bounded LayerNorm affine factors and bounded zeroth biases (ablation)ViT · ImageNet-1K pretrainingtraining sparsityImproved

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