Signals 4 · free daily AI digest

A Spectral Theory of Grokking: Weight Decay induces Feature Learning

Paper recorded by Signals 4 on 2026-09-22 in cs.AI. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-22 on arXiv · recorded by Signals 4 on 2026-09-23

Category: cs.AI · 人工智能 · first seen 2026-09-23

Abstract

In grokking an early fit to the training data separates from a much later improvement in generalization. During this delay, training can move from a fixed neural tangent kernel (NTK) regime to one in which task-relevant kernel eigendirections continue to evolve. We provide a quantitative theory for how this transition from lazy to rich learning can produce delayed generalization. For homogeneous n

Read on arXiv →

#16 most recent of 360 cs.AI papers we have recorded · ↑ newer: From Alignment to Access Control: A Framework for GenAI Policy Enforce · ↓ older: The Delegation Blind Spot: Auditing Product Decisions from Agent Choic
Cite this page: A Spectral Theory of Grokking: Weight Decay induces Feature Learning: the #16 most recent of 360 cs.AI papers we have recorded (as of 2026-09-22). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/a-spectral-theory-of-grokking-weight-decay-induces-feature-learning.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.AI papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions