English
Related papers

Related papers: Pre-trained Models Perform the Best When Token Dis…

200 papers

The problem of compression in standard information theory consists of assigning codes as short as possible to numbers. Here we consider the problem of optimal coding -- under an arbitrary coding scheme -- and show that it predicts Zipf's…

Computation and Language · Computer Science 2020-09-24 Ramon Ferrer-i-Cancho , Christian Bentz , Caio Seguin

When following a sequence - such as reading a text or tracking a user's activity - one can measure how the "dictionary" of distinct elements (types) grows with the number of observations (tokens). When this growth follows a power law, it is…

Physics and Society · Physics 2026-04-21 Célestin Zimmerlin , Thomas Louail , Manuel Moussallam , Marc Barthelemy

The binary many-step Markov chain with the step-like memory function is considered as a model for the analysis of rank distributions of words in stochastic symbolic dynamical systems. We prove that the envelope curve for this distribution…

History and Philosophy of Physics · Physics 2007-05-23 K. E. Kechedzhy O. V. Usatenko , V. A. Yampol'skii

Past work has established scaling laws that predict the performance of a neural language model (LM) as a function of its parameter count and the number of tokens it's trained on, enabling optimal allocation of a fixed compute budget. Are…

Computation and Language · Computer Science 2024-05-28 Rohan Pandey

Traditional greedy tokenization methods have been a critical step in Natural Language Processing (NLP), influencing how text is converted into tokens and directly impacting model performance. While subword tokenizers like Byte-Pair Encoding…

Computation and Language · Computer Science 2025-05-05 Bharath Raj , Garvit Suri , Vikrant Dewangan , Raghav Sonavane

In this paper we combine statistical analysis of large text databases and simple stochastic models to explain the appearance of scaling laws in the statistics of word frequencies. Besides the sublinear scaling of the vocabulary size with…

Physics and Society · Physics 2014-11-05 Martin Gerlach , Eduardo G. Altmann

We introduce a scaling law for fine-tuning large language models (LLMs) under fixed compute budgets that explicitly accounts for data composition. Conventional approaches measure training data solely by total tokens, yet the number of…

Computation and Language · Computer Science 2025-06-04 Ryan Lagasse , Aidan Kierans , Avijit Ghosh , Shiri Dori-Hacohen

Tokenization is associated with many poorly understood shortcomings in language models (LMs), yet remains an important component for long sequence scaling purposes. This work studies how tokenization impacts model performance by analyzing…

Computation and Language · Computer Science 2025-04-15 Buu Phan , Brandon Amos , Itai Gat , Marton Havasi , Matthew Muckley , Karen Ullrich

Bayesian modelling and statistical text analysis rely on informed probability priors to encourage good solutions. This paper empirically analyses whether text in medical discharge reports follow Zipf's law, a commonly assumed statistical…

Computation and Language · Computer Science 2020-11-09 Juan C Quiroz , Liliana Laranjo , Catalin Tufanaru , Ahmet Baki Kocaballi , Dana Rezazadegan , Shlomo Berkovsky , Enrico Coiera

Neural scaling laws define a predictable relationship between a model's parameter count and its performance after training in the form of a power law. However, most research to date has not explicitly investigated whether scaling laws can…

Computation and Language · Computer Science 2022-10-19 Maor Ivgi , Yair Carmon , Jonathan Berant

We present frequency-ordered tokenization, a simple preprocessing technique that improves lossless text compression by exploiting the power-law frequency distribution of natural language tokens (Zipf's law). The method tokenizes text with…

Information Theory · Computer Science 2026-02-27 Maximilian Kalcher

Language model (LM) decoding is based on the next-token prediction (NTP) probability distribution. For neural LMs (e.g., Transformer-based), NTP distribution is essentially a softmax-regularized dot product between an encoded input context…

Computation and Language · Computer Science 2024-10-04 Letian Peng , Chenyang An , Jingbo Shang

Zipf's law predicts a power-law relationship between word rank and frequency in language communication systems and has been widely reported in a variety of natural language processing applications. However, the emergence of natural language…

Computation and Language · Computer Science 2018-12-05 Bohdan Khomtchouk , Shyam Sudhakaran

Tokenization significantly influences language models(LMs)' performance. This paper traces the evolution of tokenizers from word-level to subword-level, analyzing how they balance tokens and types to enhance model adaptability while…

Computation and Language · Computer Science 2024-03-04 Jinbiao Yang

The Zipf's law establishes that if the words of a (large) text are ordered by decreasing frequency, the frequency versus the rank decreases as a power law with exponent close to $-1$. Previous work has stressed that this pattern arises from…

Physics and Society · Physics 2019-04-03 Felipe Urbina , Javier Vera

We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven…

Scaling laws enable the optimal selection of data amount and language model size, yet the impact of the data unit, the token, on this relationship remains underexplored. In this work, we systematically investigate how the information…

Computation and Language · Computer Science 2026-05-27 Tomasz Limisiewicz , Artidoro Pagnoni , Srini Iyer , Mike Lewis , Sachin Mehta , Alisa Liu , Margaret Li , Gargi Ghosh , Luke Zettlemoyer

Subword tokenization is a key part of many NLP pipelines. However, little is known about why some tokenizer and hyperparameter combinations lead to better downstream model performance than others. We propose that good tokenizers lead to…

Computation and Language · Computer Science 2023-06-30 Vilém Zouhar , Clara Meister , Juan Luis Gastaldi , Li Du , Mrinmaya Sachan , Ryan Cotterell

Despite the recent successes of deep learning in natural language processing (NLP), there remains widespread usage of and demand for techniques that do not rely on machine learning. The advantage of these techniques is their…

Computation and Language · Computer Science 2020-12-04 Adam Hare , Yu Chen , Yinan Liu , Zhenming Liu , Christopher G. Brinton

We checked that the distribution of words in text should uniform, which gives Heaps' law as natural result, that is, the number of types of words can be expressed as a power law of the number of tokens within text. We developed a…

Physics and Society · Physics 2025-04-16 Kim Chol-jun