---
title: "At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics — Stuff That Spins"
description: "arXiv:2607.06639v1 Announce Type: new Abstract: On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has alre…"
	canonical: "https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics"
html: "https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics"
json: "https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics.json"
markdown: "https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics.md"
keywords: ["narrative intelligence", "SpinGraph", "AI recall"]
date: "2026-07-09T04:00:00+00:00"
modified: "2026-07-09T06:04:54.41143+00:00"
json_ld: |
  {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://stuffthatspins.com/#organization","name":"Stuff That Spins","url":"https://stuffthatspins.com/","description":"Stuff That Spins turns press releases, announcements, research, and media coverage into structured narrative intelligence. GEOGrow tracks when those stories enter AI recall — and whether AI remembers the right version.","logo":{"@type":"ImageObject","url":"https://stuffthatspins.com/images/logo.png"},"sameAs":[]},{"@type":"NewsArticle","@id":"https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics#article","headline":"At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics","description":"arXiv:2607.06639v1 Announce Type: new Abstract: On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has alre…","datePublished":"2026-07-09T04:00:00+00:00","dateModified":"2026-07-09T06:04:54.41143+00:00","url":"https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics","mainEntityOfPage":{"@type":"WebPage","@id":"https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics"},"isAccessibleForFree":true,"inLanguage":"en-US","articleSection":"research","author":{"@type":"Organization","name":"arXiv Machine Learning","url":"https://export.arxiv.org/rss/cs.LG"},"publisher":{"@id":"https://stuffthatspins.com/#organization"},"citation":"https://arxiv.org/abs/2607.06639","about":[],"mentions":[{"@type":"Organization","name":"arXiv Machine Learning"}]},{"@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Stuff That Spins","item":"https://stuffthatspins.com/"},{"@type":"ListItem","position":2,"name":"At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics","item":"https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics"}]}]}
---

# At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics

**Source:** Unknown  
**Published:** July 9, 2026  
**Original:** https://arxiv.org/abs/2607.06639  

## On this page

- [Overview](#overview)

<a id="overview"></a>

## Overview

arXiv:2607.06639v1 Announce Type: new Abstract: On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has already generalized. Reading effective rank at the grokking transition overstates the converged value by 3-5x on an MLP, and by 1.3-1.5x on a transformer trained to convergence; on the MLP it also erases which cells compress at all. Compression lags the accuracy transition by an amount on the order of the time-to-grok, at least 10,000 steps, r

---
*HTML version: https://stuffthatspins.com/spin/at-grok-is-not-convergeda-measurement-validity-audit-for-grokking-representation-metrics*