Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training

• URL: http://arxiv.org/abs/2508.15390v2
• Date: Mon, 27 Oct 2025 02:39:13 GMT
• Title: Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training
• Authors: Woojin Chung, Jeonghoon Kim,
• Abstract summary: Large language models are trained with tokenizers, and the resulting token distribution is highly imbalanced.<n>Recent practice favors ever-larger vocabularies, but it is unclear where the benefit comes from.<n>We show that larger vocabularies reduce this complexity.
• Score: 10.990131879961261
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Large language models are trained with tokenizers, and the resulting token distribution is highly imbalanced: a few words dominate the stream while most occur rarely. Recent practice favors ever-larger vocabularies, but it is unclear where the benefit comes from. To this end, we perform a controlled study that scales the vocabulary of the language model from 24K to 196K while holding data, computation, and optimization unchanged. We begin by quantifying the complexity of tokenized text -- formalized via Kolmogorov complexity -- and show that larger vocabularies reduce this complexity. Above 24K, every common word is already tokenized as a single token, so enlarging vocabulary only deepens the relative token-frequency imbalance. Word-level loss decomposition shows that larger vocabularies reduce cross-entropy loss almost exclusively by lowering uncertainty on the 2,500 most frequent words, even though loss on the rare tail rises. The same frequent words cover roughly 75% of tokens in downstream benchmarks, so this training advantage transfers intact. We further show that enlarging model parameters with a fixed vocabulary yields the same frequent-word benefit. Our results recast "bigger vocabularies help" as "lowering complexity of tokenized text helps," offering a simple, principled knob for tokenizer--model co-design and clarifying the loss dynamics that govern language model scaling in pre-training.

Related papers

• LiteToken: Removing Intermediate Merge Residues From BPE Tokenizers [76.59130257385826]
  Intermediate merge residues in BPE vocabularies are frequent during merge learning so that retained in the final vocabulary, but are mostly further merged and rarely emitted when tokenizing the corpus during tokenizer usage.<n>We present a systematic empirical characterization of this phenomenon across commonly used tokenizers and introduce LiteToken, a simple method for removing residue tokens.<n>Experiments show that LiteToken reduces token fragmentation, reduces parameters, and improves robustness to noisy or misspelled inputs, while preserving overall performance.
  arXiv  Detail & Related papers  (2026-02-04T16:19:05Z)
• Training Language Models with homotokens Leads to Delayed Overfitting [2.531076482407163]
  Subword tokenization introduces a computational layer in language models where many distinct token sequences decode to the same surface form and preserve meaning.<n>We formalize homotoken-as a strictly meaning-preserving form of data augmentation.<n>In data-constrained pretraining, homotoken augmentation consistently delays overfitting under repeated data exposure.<n>In multilingual fine-tuning, we find that the effectiveness of homotokens depends on tokenizer quality.
  arXiv  Detail & Related papers  (2026-01-06T09:57:00Z)
• Lossless Vocabulary Reduction for Auto-Regressive Language Models [21.015330660860865]
  Tokenization -- the process of decomposing a given text into a sequence of subwords called tokens -- is one of the key components in the development of language models.<n>We establish a theoretical framework of lossless vocabulary reduction, which efficiently converts a given auto-regressive language model into the one with an arbitrarily small vocabulary.<n>As an application, we demonstrate that language models with different tokenization can cooperate with each other efficiently through their maximal common vocabulary.
  arXiv  Detail & Related papers  (2025-10-09T11:38:48Z)
• Broken Tokens? Your Language Model can Secretly Handle Non-Canonical Tokenizations [83.93566096400723]
  We find that instruction-tuned models retain up to 93.4% of their original performance when given a randomly sampled tokenization.<n>Character-level segmentation improves string manipulation and code understanding tasks by up to +14%.<n>Right-aligned digit grouping enhances large-number arithmetic by +33%.
  arXiv  Detail & Related papers  (2025-06-23T18:02:26Z)
• Causal Estimation of Tokenisation Bias [58.20086589761273]
  We quantify the effect of including or not a subword in a tokeniser's vocabulary on the probability a trained model assigns to the corresponding characters.<n>We find that tokenisation consistently affects models' outputs across scales, vocabularies, and tokenisers.<n> Notably, a subword's presence in a small model's vocabulary may increase its characters' probability by up to 17 times.
  arXiv  Detail & Related papers  (2025-06-03T17:59:47Z)
• An Analysis of BPE Vocabulary Trimming in Neural Machine Translation [56.383793805299234]
  vocabulary trimming is a postprocessing step that replaces rare subwords with their component subwords. We show that vocabulary trimming fails to improve performance and is even prone to incurring heavy degradation.
  arXiv  Detail & Related papers  (2024-03-30T15:29:49Z)
• MiLe Loss: a New Loss for Mitigating the Bias of Learning Difficulties in Generative Language Models [40.992566245706996]
  We propose a MiLe Loss function for mitigating the bias of learning difficulties with tokens. We train generative language models at different scales of 468M, 1.2B, and 6.7B parameters. Experiments reveal that models incorporating the proposed MiLe Loss can gain consistent performance improvement on downstream benchmarks.
  arXiv  Detail & Related papers  (2023-10-30T13:33:21Z)
• Learn Your Tokens: Word-Pooled Tokenization for Language Modeling [11.40976202290724]
  Language models typically tokenize text into subwords, using a deterministic, hand-engineered of combining tokens into longer strings. Recent attempts to compress and limit context lengths with fixed size convolutions is helpful but completely ignores the word boundary. This paper considers an alternative 'learn your word' scheme which utilizes the word boundary to pool bytes/characters into word representations.
  arXiv  Detail & Related papers  (2023-10-17T23:34:39Z)
• CompoundPiece: Evaluating and Improving Decompounding Performance of Language Models [77.45934004406283]
  We systematically study decompounding, the task of splitting compound words into their constituents. We introduce a dataset of 255k compound and non-compound words across 56 diverse languages obtained from Wiktionary. We introduce a novel methodology to train dedicated models for decompounding.
  arXiv  Detail & Related papers  (2023-05-23T16:32:27Z)
• ngram-OAXE: Phrase-Based Order-Agnostic Cross Entropy for Non-Autoregressive Machine Translation [51.06378042344563]
  A new training oaxe loss has proven effective to ameliorate the effect of multimodality for non-autoregressive translation (NAT) We extend oaxe by only allowing reordering between ngram phrases and still requiring a strict match of word order within the phrases. Further analyses show that ngram-oaxe indeed improves the translation of ngram phrases, and produces more fluent translation with a better modeling of sentence structure.
  arXiv  Detail & Related papers  (2022-10-08T11:39:15Z)
• Quark: Controllable Text Generation with Reinforced Unlearning [68.07749519374089]
  Large-scale language models often learn behaviors that are misaligned with user expectations. We introduce Quantized Reward Konditioning (Quark), an algorithm for optimizing a reward function that quantifies an (un)wanted property. For unlearning toxicity, negative sentiment, and repetition, our experiments show that Quark outperforms both strong baselines and state-of-the-art reinforcement learning methods.
  arXiv  Detail & Related papers  (2022-05-26T21:11:51Z)
• How Robust is Neural Machine Translation to Language Imbalance in Multilingual Tokenizer Training? [86.48323488619629]
  We analyze how translation performance changes as the data ratios among languages vary in the tokenizer training corpus. We find that while relatively better performance is often observed when languages are more equally sampled, the downstream performance is more robust to language imbalance than we usually expected.
  arXiv  Detail & Related papers  (2022-04-29T17:50:36Z)
• A Vocabulary-Free Multilingual Neural Tokenizer for End-to-End Task Learning [8.052271364177988]
  Subword tokenization is a commonly used input pre-processing step in most recent NLP models. We propose a vocabulary-free neural tokenizer by distilling segmentation information from subword tokenization. Our tokenizer consistently improves performance on multilingual (NLI) and code-switching (sentiment analysis) tasks.
  arXiv  Detail & Related papers  (2022-04-22T16:50:49Z)

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.