# The Inverse Scaling Effect of Pre-Trained Language Model Surprisal Is Not Due to Data Leakage

arXiv: https://arxiv.org/abs/2506.01172 · HTML: https://mlsanity.com/paper/2506.01172/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 13 of its 43 references that are in the map.
- First posted on arXiv 1 Jun 2025.

## Builds on (most cited first, 13 of 13)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [ZeRO: Memory Optimizations Toward Training Trillion Parameter Models](https://mlsanity.com/paper/1910.02054/) (2019)
- [Why Does Surprisal From Larger Transformer-Based Language Models Provide a Poorer Fit to Human Reading Times?](https://mlsanity.com/paper/2212.12131/) (2022)
- [Syntactic Structure from Deep Learning](https://mlsanity.com/paper/2004.10827/) (2020)
- [Testing the Predictions of Surprisal Theory in 11 Languages](https://mlsanity.com/paper/2307.03667/) (2023)
- [Transformer-Based Language Model Surprisal Predicts Human Reading Times Best with About Two Billion Training Tokens](https://mlsanity.com/paper/2304.11389/) (2023)
- [The Natural Stories Corpus](https://mlsanity.com/paper/1708.05763/) (2017)
- [Frequency Explains the Inverse Correlation of Large Language Models' Size, Training Data Amount, and Surprisal's Fit to Reading Times](https://mlsanity.com/paper/2402.02255/) (2024)
- [How to Compute the Probability of a Word](https://mlsanity.com/paper/2406.14561/) (2024)
- [Leading Whitespaces of Language Models' Subword Vocabulary Pose a Confound for Calculating Word Probabilities](https://mlsanity.com/paper/2406.10851/) (2024)
- [Evaluating n-Gram Novelty of Language Models Using Rusty-DAWG](https://mlsanity.com/paper/2406.13069/) (2024)

## Built on by (most cited first, 5 of 5)

- [N-gram-like Language Models Predict Naturalistic Reading Time Best](https://mlsanity.com/paper/2603.09872/) (2026)
- [On the scaling relationship between cloze probabilities and language model next-token prediction](https://mlsanity.com/paper/2602.17848/) (2026)
- [Surprisal Theory is Tautological (without Rational Grounding)](https://mlsanity.com/paper/2607.21574/) (2026)
- [A Systematic Analysis of the Predictive Power of LM Surprisal in Reading Chinese](https://mlsanity.com/paper/2610.04898/) (2026)
- [Estimating Surprisal from the Speech Signal](https://mlsanity.com/paper/2610.18439/) (2026)

## Abstract (arXiv)

In psycholinguistic modeling, surprisal from larger pre-trained language models has been shown to be a poorer predictor of naturalistic human reading times. However, it has been speculated that this may be due to data leakage that caused language models to see the text stimuli during training. This paper presents two studies to address this concern at scale. The first study reveals relatively little leakage of five naturalistic reading time corpora in two pre-training datasets in terms of length and frequency of token n-gram overlap. The second study replicates the negative relationship between language model size and the fit of surprisal to reading times using models trained on 'leakage-free' data that overlaps only minimally with the reading time corpora. Taken together, this suggests that previous results using language models trained on these corpora are not driven by the effects of data leakage.