# Grokking in LLM Pretraining? Monitor Memorization-to-Generalization without Test

arXiv: https://arxiv.org/abs/2506.21551 · HTML: https://mlsanity.com/paper/2506.21551/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in MoE large language models, one of 12 directions in Mixture-of-experts.
- Builds on 25 of its 32 references that are in the map.
- First posted on arXiv 26 Jun 2025.

## Placed in

- Mixture-of-experts / [MoE large language models](https://mlsanity.com/moe/moe-large-language-models/)

## Builds on (most cited first, 25 of 25)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering](https://mlsanity.com/paper/1809.02789/) (2018)
- [Progress measures for grokking via mechanistic interpretability](https://mlsanity.com/paper/2301.05217/) (2023)
- [Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets](https://mlsanity.com/paper/2201.02177/) (2022)
- [Towards Reasoning in Large Language Models: A Survey](https://mlsanity.com/paper/2212.10403/) (2022)
- [Towards Understanding Grokking: An Effective Theory of Representation Learning](https://mlsanity.com/paper/2205.10343/) (2022)
- [Omnigrok: Grokking Beyond Algorithmic Data](https://mlsanity.com/paper/2210.01117/) (2022)
- [Min-K%++: Improved Baseline for Detecting Pre-Training Data from Large Language Models](https://mlsanity.com/paper/2404.02936/) (2024)
- [A Tale of Two Circuits: Grokking as Competition of Sparse and Dense Subnetworks](https://mlsanity.com/paper/2303.11873/) (2023)
- [Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization](https://mlsanity.com/paper/2405.15071/) (2024)
- [Deep Networks Always Grok and Here is Why](https://mlsanity.com/paper/2402.15555/) (2024)
- [A Closer Look into Mixture-of-Experts in Large Language Models](https://mlsanity.com/paper/2406.18219/) (2024)
- [Grokking at the Edge of Numerical Stability](https://mlsanity.com/paper/2501.04697/) (2025)
- [Predicting Grokking Long Before it Happens: A look into the loss landscape of models which grok](https://mlsanity.com/paper/2306.13253/) (2023)
- [Unified View of Grokking, Double Descent and Emergent Abilities: A Perspective from Circuits Competition](https://mlsanity.com/paper/2402.15175/) (2024)
- [Understanding Grokking Through A Robustness Viewpoint](https://mlsanity.com/paper/2311.06597/) (2023)
- [Dual Instruction Tuning with Large Language Models for Mathematical Reasoning](https://mlsanity.com/paper/2403.18295/) (2024)

## Built on by (most cited first, 10 of 10)

- [Rote Learning Considered Useful: Generalizing over Memorized Data in LLMs](https://mlsanity.com/paper/2507.21914/) (2025)
- [Rewiring Experts on the Fly:Continuous Rerouting for Better Online Adaptation in Mixture-of-Expert models](https://mlsanity.com/paper/2510.14853/) (2025)
- [Spectral Entropy Collapse as a Phase Transition in Delayed Generalisation: An Interventional and Predictive Framework for Grokkin](https://mlsanity.com/paper/2604.13123/) (2026)
- [Opening the Black Box: A Survey on the Mechanisms of Multi-Step Reasoning in Large Language Models](https://mlsanity.com/paper/2601.14270/) (2026)
- [Beyond Temperature: Hyperfitting as a Late-Stage Geometric Expansion](https://mlsanity.com/paper/2605.22579/) (2026)
- [A Pre-Training Analogue of Grokking in Language Models: Tracing Delayed Grammatical Generalization](https://mlsanity.com/paper/2606.00230/) (2026)
- [Train Smarter, Not Longer: Memorization-Guided Data Reuse for Efficient LLM Training](https://mlsanity.com/paper/2607.04969/) (2026)
- [Deep Exploration of Epoch-wise Double Descent in Noisy Data: Signal Separation, Large Activation, and Benign Overfitting](https://mlsanity.com/paper/2601.08316/) (2026)
- [Neuro-symbolic Syntactic Parsing: Shaping a Neural Network with the CYK Algorithm](https://mlsanity.com/paper/2605.31421/) (2026)
- [Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in Large Language Model Finetuning](https://mlsanity.com/paper/2607.08393/) (2026)

## Abstract (arXiv)

This paper presents the first study of grokking in practical LLM pretraining. Specifically, we investigate when an LLM memorizes the training data, when its generalization on downstream tasks starts to improve, and what happens if there is a lag between the two. Unlike existing works studying when a small model generalizes to limited and specified tasks during thousands epochs' training on algorithmic data, we focus on a practical setting for LLMs, i.e., one-epoch pretraining of next-token prediction on a cross-domain, large-scale corpus, and generalization on diverse benchmark tasks covering math/commonsense reasoning, code generation, and domain-specific retrieval. Our study, for the first time, verifies that grokking still emerges in pretraining mixture-of-experts (MoE) LLMs, though different local data groups may enter their grokking stages asynchronously due to the heterogeneity of their distributions and attributions to others. To find a mechanistic interpretation of this local grokking, we investigate the dynamics of training data's pathways (i.e., expert choices across layers in MoE). Our primary discovery is that the pathways evolve from random, non-smooth across layers, instance-specific to more structured and transferable across samples, despite the converged pretraining loss. This depicts a transition from memorization to generalization. Two novel metrics are developed to quantify these patterns: one computes the pathway similarity between samples, while the other measures the consistency of aggregated experts between subsequent layers for each sample. These training data based metrics induce zero cost but can faithfully track and monitor the generalization of LLMs on downstream tasks, which, in conventional settings, requires costly instruction tuning and benchmark evaluation.