On Linear Representations and Pretraining Data Frequency in Language Models

• URL: http://arxiv.org/abs/2504.12459v1
• Date: Wed, 16 Apr 2025 19:50:03 GMT
• Title: On Linear Representations and Pretraining Data Frequency in Language Models
• Authors: Jack Merullo, Noah A. Smith, Sarah Wiegreffe, Yanai Elazar,
• Abstract summary: We study the connection between pretraining data frequency and models' linear representations of factual relations.<n>We find evidence that the formation of linear representations is strongly connected to pretraining term frequencies.<n>We conclude that the strength of linear representations in LMs contains signal about the models' pretraining corpora.
• Score: 54.756179696806356
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Pretraining data has a direct impact on the behaviors and quality of language models (LMs), but we only understand the most basic principles of this relationship. While most work focuses on pretraining data's effect on downstream task behavior, we investigate its relationship to LM representations. Previous work has discovered that, in language models, some concepts are encoded `linearly' in the representations, but what factors cause these representations to form? We study the connection between pretraining data frequency and models' linear representations of factual relations. We find evidence that the formation of linear representations is strongly connected to pretraining term frequencies; specifically for subject-relation-object fact triplets, both subject-object co-occurrence frequency and in-context learning accuracy for the relation are highly correlated with linear representations. This is the case across all phases of pretraining. In OLMo-7B and GPT-J, we discover that a linear representation consistently (but not exclusively) forms when the subjects and objects within a relation co-occur at least 1k and 2k times, respectively, regardless of when these occurrences happen during pretraining. Finally, we train a regression model on measurements of linear representation quality in fully-trained LMs that can predict how often a term was seen in pretraining. Our model achieves low error even on inputs from a different model with a different pretraining dataset, providing a new method for estimating properties of the otherwise-unknown training data of closed-data models. We conclude that the strength of linear representations in LMs contains signal about the models' pretraining corpora that may provide new avenues for controlling and improving model behavior: particularly, manipulating the models' training data to meet specific frequency thresholds.

Related papers

• Tessellated Linear Model for Age Prediction from Voice [29.0093388690853]
  Tessellated Linear Model (TLM) is a piecewise linear approach that combines the simplicity of linear models with the capacity of non-linear functions.<n>We evaluate TLM on the TIMIT dataset on the task of age prediction from voice, where it outperformed state-of-the-art deep learning models.
  arXiv  Detail & Related papers  (2025-01-16T01:28:45Z)
• Causal Estimation of Memorisation Profiles [58.20086589761273]
  Understanding memorisation in language models has practical and societal implications. Memorisation is the causal effect of training with an instance on the model's ability to predict that instance. This paper proposes a new, principled, and efficient method to estimate memorisation based on the difference-in-differences design from econometrics.
  arXiv  Detail & Related papers  (2024-06-06T17:59:09Z)
• Measuring and Improving Attentiveness to Partial Inputs with Counterfactuals [91.59906995214209]
  We propose a new evaluation method, Counterfactual Attentiveness Test (CAT) CAT uses counterfactuals by replacing part of the input with its counterpart from a different example, expecting an attentive model to change its prediction. We show that GPT3 becomes less attentive with an increased number of demonstrations, while its accuracy on the test data improves.
  arXiv  Detail & Related papers  (2023-11-16T06:27:35Z)
• Self-Influence Guided Data Reweighting for Language Model Pre-training [46.57714637505164]
  Language Models (LMs) pre-trained with self-supervision on large text corpora have become the default starting point for developing models for various NLP tasks. All data samples in the corpus are treated with equal importance during LM pre-training. Due to varying levels of relevance and quality of data, equal importance to all the data samples may not be the optimal choice. We propose PRESENCE, a method for jointly reweighting samples by leveraging self-influence (SI) scores as an indicator of sample importance and pre-training.
  arXiv  Detail & Related papers  (2023-11-02T01:00:46Z)
• Stubborn Lexical Bias in Data and Models [50.79738900885665]
  We use a new statistical method to examine whether spurious patterns in data appear in models trained on the data. We apply an optimization approach to *reweight* the training data, reducing thousands of spurious correlations. Surprisingly, though this method can successfully reduce lexical biases in the training data, we still find strong evidence of corresponding bias in the trained models.
  arXiv  Detail & Related papers  (2023-06-03T20:12:27Z)
• Analysis of Interpolating Regression Models and the Double Descent Phenomenon [3.883460584034765]
  It is commonly assumed that models which interpolate noisy training data are poor to generalize. The best models obtained are overparametrized and the testing error exhibits the double descent behavior as the model order increases. We derive a result based on the behavior of the smallest singular value of the regression matrix that explains the peak location and the double descent shape of the testing error as a function of model order.
  arXiv  Detail & Related papers  (2023-04-17T09:44:33Z)
• Impact of Pretraining Term Frequencies on Few-Shot Reasoning [51.990349528930125]
  We investigate how well pretrained language models reason with terms that are less frequent in the pretraining data. We measure the strength of this correlation for a number of GPT-based language models on various numerical deduction tasks. Although LMs exhibit strong performance at few-shot numerical reasoning tasks, our results raise the question of how much models actually generalize beyond pretraining data.
  arXiv  Detail & Related papers  (2022-02-15T05:43:54Z)
• On the Transferability of Pre-trained Language Models: A Study from Artificial Datasets [74.11825654535895]
  Pre-training language models (LMs) on large-scale unlabeled text data makes the model much easier to achieve exceptional downstream performance. We study what specific traits in the pre-training data, other than the semantics, make a pre-trained LM superior to their counterparts trained from scratch on downstream tasks.
  arXiv  Detail & Related papers  (2021-09-08T10:39:57Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.