AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

The Geometric Origin of Grokking: Accelerating Generalization via Active Structural Reorganization

Kefei Tao, Zhang Zhang, Mingze Qi, Xiaojun Duan

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory

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 R2G losstransformer baseline · P=500 modular additionruns reaching full grokking: +73%Improved
add R2G losstransformer baseline · P=499 modular additiongrokking rate: +80%Improved
add R2G losstransformer baseline · P=113 modular multiplicationgrokking rate: +67%Improved
add R2G loss2-layer decoder-only Transformer · P=113 modular additiongrokking rate: +80%Improved
add R2G losstransformer baseline · Tense-Inflectiongrokking success rate: +67%Improved
use R2G instead of GrokfastP=500 modular addition · P=500 modular additiongrokking success rate: 73% vs 60%Improved
use R2G instead of NeuralGrokP=500 modular addition · P=500 modular additiongrokking success rate: 73% vs 27%Improved
add angular term to norm-only loss (full R2G) (ablation)R2G loss · non-grokking datasetsgrokking ratio: 45% -> 65%Improved
add norm term to angular-only loss (full R2G) (ablation)R2G loss · non-grokking datasetsgrokking ratio: 55% -> 65%; mean epoch 2500 -> 1882Improved
use angular-only instead of norm-only loss (ablation)decoupled R2G variants · non-grokking datasetsgrokking ratio: 45% -> 55%Improved
use reciprocal R2G form instead of LayerNorm objective (ablation)R2G loss · P=500 modular additiongrokking ratio: 60% -> 73%Improved
use reciprocal R2G form instead of Hyperball objective (ablation)R2G loss · P=500 modular additiongrokking ratio: 33% -> 73%Improved

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