A Comparative Analysis of LLM Memorization at Statistical and Internal Levels: Cross-Model Commonalities and Model-Specific Signatures

Bowen Chen; Namgi Han; Yusuke Miyao

cs.CLMar 23, 2026

A Comparative Analysis of LLM Memorization at Statistical and Internal Levels: Cross-Model Commonalities and Model-Specific Signatures

Authors: Bowen Chen, Namgi Han, Yusuke Miyao

Organizations: Department of Computer Science, The University of Tokyo · Research and Development Center for Large Language Models, National Institute of Informatics

Abstract

Memorization is a fundamental component of intelligence for both humans and LLMs. However, while LLM performance scales rapidly, our understanding of memorization lags. Due to limited access to the pre-training data of LLMs, most previous studies focus on a single model series, leading to isolated observations among series, making it unclear which findings are general or specific. In this study, we collect multiple model series (Pythia, OpenLLaMa, StarCoder, OLMo1/2/3) and analyze their shared or unique memorization behavior at both the statistical and internal levels, connecting individual observations while showing new findings. At the statistical level, we reveal that the memorization rate scales log-linearly with model size, and memorized sequences can be further compressed. Further analysis demonstrated a shared frequency and domain distribution pattern for memorized sequences. However, different models also show individual features under the above observations. At the internal level, we find that LLMs can remove certain injected perturbations, while memorized sequences are more sensitive. By decoding middle layers and attention head ablation, we revealed the general decoding process and shared important heads for memorization. However, the distribution of those important heads differs between families, showing a unique family-level feature. Through bridging various experiments and revealing new findings, this study paves the way for a universal and fundamental understanding of memorization in LLM.

Figures & tables

Model	Size	Pre-training Data
Pythia	160m, 410m, 1b	Pile Gao et al. (2020)
Pythia	2.8b, 6.9b, 12b	Pile Gao et al. (2020)
OLMo1	1b, 7b	Dolma Soldaini et al. (2024)
OLMo2	1b, 7b, 13b	OLMo-Mix and Domino
OLMo2	32b	OLMo et al. (2025)
OLMo3	7b, 32b	Dolma3 Olmo et al. (2025)

Table 1: Model families and their pre-training data.

Panel of Figure 1 — Figure 1: (Top Figure) The relationship between memorization rate, model size, and pre-training corpus size. The x/y-axis represents parameter counts and memorization rate on a log scale. Line width and color also encode the training token count (thicker lines = more tokens). (Bottom left) The compression ratio across models. A smaller compression ratio requires fewer tokens to generate the same memorized sequences. (Bottom right) The probability density distribution of memorization scores for all models (log-scaled y-axis).

Figure 2: The average token frequency distribution for different models. The upper figure shows the normalized frequency for memorized sequences. The dotted vertical lines indicate the average frequency for each model. The bottom figure shows the absolute frequency for memorized (solid line) and unmemorized (dotted line) sequences.

Model	Pythia			OLMo1		OLMo2			OLMo3		OpenLLaMA
Size	1B	2.8B	12B	1B	7B	7B	13B	32B	7B	32B	3B	7B	13B
Structural	75.1%	76.3%	74.3%	57.1%	49.7%	44.7%	43.9%	39.8%	49.0%	44.4%	65.9%	61.3%	58.7%
Semi-Structural	19.4%	17.2%	17.1%	32.3%	34.6%	9.7%	10.5%	12.1%	6.4%	7.9%	12.1%	13.5%	13.2%
Free-Text	5.5%	6.5%	8.6%	10.6%	15.7%	45.6%	45.6%	48.1%	44.7%	47.7%	21.9%	25.2%	28.1%

Table 2: Proportion of memorized texts across Structural, Semi-Structural, and Free-Text categories. StarCoder is not presented as it is only trained on code.

Panel of Figure 3 — Figure 3: (Left figure) Memorization score distribution shift under noise. (Right figure) The Attention Head similarity between clean and noised models, with variance represented by the shadow. Solid/dotted lines represent results for memorized/unmemorized sequences. The noise range is [0.1, 0.2, 0.3, 0.4, 0.5].

Figure 4: Decoding probability (Y-axis) for memorized and unmemorized tokens across different layers (X-axis) for all models. The solid/dotted line represents the probability for memorized/unmemorized sequences.

Model	Size	Any Two Domain Pair			In-Domain Example Pair	Percentage of Shared Important Heads Among $>x\%$ Domains
		Mean Jaccard	Max Jaccard	Random	Mean Jaccard	$>30\%$ dom	$>50\%$ dom	$>70\%$ dom	$>90\%$ dom	all dom
OLMo1	1b	28.58%	57.58%	11.39%	57.86%	28.12%	12.11%	5.08%	1.17%	0.78%
OLMo1	7b	24.41%	48.01%	11.14%	80.03%	29.79%	10.84%	2.05%	0.98%	0.98%
OLMo2	1b	52.08%	89.09%	11.39%	57.69%	23.05%	18.75%	14.06%	7.42%	4.69%
	7b	21.97%	70.12%	11.14%	82.26%	23.24%	9.96%	3.42%	0.78%	0.29%
	13b	18.93%	38.23%	11.12%	87.66%	20.62%	8.94%	1.75%	0.62%	0.38%

Table 3: Cross-domain top-20% head overlap against a random-selection, together with mean within-domain example-level overlap and the ratio of heads that remain top-20% important in at least X%, or all domains.

Figure 5: Distribution of shared important heads across layers for Pythia-12b and OLMo-13b

Figure 6: Distribution Similarity Heatmap of Layer-wise Average Head Importance for all models.

Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Model Family	Domain Type	Domains	#Dom
Pythia	Structural	USPTO_Backgrounds, EuroParl, Enron_Emails, Github, StackExchange	5
	Semi-Structural	PubMed_Abstracts, PubMed_Central, FreeLaw, NIH_ExPorter	4
	Free-Text	ArXiv, DM_Mathematics, Gutenberg (PG-19), HackerNews, Pile-CC,	8
	Free-Text	PhilPapers, Ubuntu_IRC, Wikipedia (en)	8
OLMo1	Structural	starcoder, proof_pile_2_algebraic_stack, proof_pile_2_open_web_math	3
	Semi-Structural	redpajama_stackexchange, cc_news_head, cc_news_middle, cc_news_tail,	8

Appendix

Table 4: Complete domain classification and counts by model family. Domains are categorized as Structural (code/mathematics), Semi-Structural (mixed format), and Free-Text (natural language prose). The #Dom column shows the count of domains in each category.

Table 5: Proportion of memorized texts (score=1) across Structural, Semi-Structural, and Free-Text categories at different noise levels.

Figure 7: MLP similarity between noised and clean models. Solid/Dotted lines represent the memorized/unmemorized sequence experiment results.

Figure 8: Frequency distribution for memorized, half-memorized, and unmemorized sequences.

Panel of Figure 9 — Figure 9: Top 20% important heads overlap for Pythia and OLMo2 models on selected domains.

Panel of Figure 10 — Figure 10: Average layer importance across layers.

Figure 11: Average norm value of residual stream for different models.

Figure 12: Layer-Normalized Averaged Importance Score for Attention Heads Across Layers.

Figure 13: Decoding probability (Y-axis) for memorized and unmemorized tokens across different layers (X-axis) for all models. The solid/dotted line represents the decoding probability across layers for memorized/unmemorized sequences.

Model	Size	Domain	Count	Share	Total
OLMo1	1B	starcoder	2,834	48.99%	5,785
		cc_en_tail	575	9.94%
		c4_filtered	367	6.34%
		books	334	5.77%
		cc_news_head	318	5.50%
		proof_pile_2_open_web_math	295	5.10%

Appendix

Table 6: Domain distribution of memorized texts across model families and sizes.

Memorization in Language Models

Memorization in language models studies how training examples and facts are stored, recalled, generalized, and exposed by model parameters.

See all 111 papers

Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models
Sep 28, 2026Florian Eichin, Philipp Mondorf, Andrei Mircea +3Memorization in Language Models Data Repetition in Language Model Training
MaiNLP, Center for Information and Language Processing, LMU Munich & Munich Center for Machine Learning (MCML), Germany · Mila – Quebec AI Institute & University of Montreal, Canada · Saarland University, Germany
A Prior-Aware Metric for Efficiently Distinguishing Memorization from Generalization in Large Language Models
Feb 21, 2026Trishita Tiwari, Ari Trachtenberg, G. Edward SuhMemorization in Language Models LLM Auditing
Cornell University · Boston University · NVIDIA, Cornell University
Understanding the Interplay between Memorization and Learning in Large Language Models
Jul 20, 2025Bishwamittra Ghosh, Soumi Das, Qinyuan Wu +4Memorization in Language Models LLM Training
Max Planck Institute for Software Systems, Germany · Boston University, USA

Sep 28, 2026J/K move · Enter open · S save

Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models

Florian Eichin, Philipp Mondorf, Andrei Mircea, Yupei Du+2

MaiNLP, Center for Information and Language Processing, LMU Munich & Munich Center for Machine Learning (MCML), Germany · Mila – Quebec AI Institute & University of Montreal, Canada · Saarland University, Germany

Memorization in Language Models Data Repetition in Language Model Training Gradient Alignment

Open paper AskPDF arXiv

Memorization has been proposed as a mechanism to explain how language models fit the tail of their training distributions, but its training dynamics are not understood well. In this work, we take a fine-grained look at memorization by decomposing the loss trajectory of memorized sequences over training and model parameters. Across the Pythia family, we study memorization of duplicated training sequences (recitation) and rare ones (recollection). We find that memorization in both cases is characterized by sequence-level gradient alignment, though recitation suffers from misalignment with other training influences which causes forgetting, explaining the necessity for higher duplication of these examples. We further show that the lower model layers are the most involved in memorization and forgetting. Predicting memorization, our decomposition improves over a cross-entropy baseline, especially in larger models and early in training. Intervening on a small set of highly influential parameters we are able to ablate memorization in the final model. Together, these findings advance our understanding of how memorization develops during training and offer insights for predicting and intervening on it.

Abstract

Memorization in Language Models

Explore similar work

Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models

A Prior-Aware Metric for Efficiently Distinguishing Memorization from Generalization in Large Language Models

Understanding the Interplay between Memorization and Learning in Large Language Models