Vocab Diet: Reshaping the Vocabulary of LLMs with Vector Arithmetic

• URL: http://arxiv.org/abs/2510.17001v1
• Date: Sun, 19 Oct 2025 20:56:58 GMT
• Title: Vocab Diet: Reshaping the Vocabulary of LLMs with Vector Arithmetic
• Authors: Yuval Reif, Guy Kaplan, Roy Schwartz,
• Abstract summary: Large language models (LLMs) encode word form variations, such as "walk"->"walked", as linear directions in embedding space.<n>Standard tokenization algorithms treat these variations as distinct tokens.<n>We propose a compact reshaping of the vocabulary: rather than assigning unique tokens to each surface form, we compose them from shared base form and transformation vectors.
• Score: 9.273273023595065
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Large language models (LLMs) were shown to encode word form variations, such as "walk"->"walked", as linear directions in embedding space. However, standard tokenization algorithms treat these variations as distinct tokens -- filling the size-capped vocabulary with surface form variants (e.g., "walk", "walking", "Walk"), at the expense of less frequent words and multilingual coverage. We show that many of these variations can be captured by transformation vectors -- additive offsets that yield the appropriate word's representation when applied to the base form word embedding -- in both the input and output spaces. Building on this, we propose a compact reshaping of the vocabulary: rather than assigning unique tokens to each surface form, we compose them from shared base form and transformation vectors (e.g., "walked" = "walk" + past tense). We apply our approach to multiple LLMs and across five languages, removing up to 10% of vocabulary entries -- thereby freeing space to allocate new, more diverse tokens. Importantly, we do so while also expanding vocabulary coverage to out-of-vocabulary words, with minimal impact on downstream performance, and without modifying model weights. Our findings motivate a foundational rethinking of vocabulary design, moving from string enumeration to a compositional vocabulary that leverages the underlying structure of language.

Related papers

• See the Text: From Tokenization to Visual Reading [63.10220471118435]
  SeeTok renders text as images (visual-text) and leverages pretrained multimodal computations to interpret them.<n>Across three different language tasks, SeeTok matches or surpasses subword tokenizers while requiring 4.43 times fewer tokens and reducing FLOPs by 70.5%.<n>SeeTok signals a shift from symbolic tokenization to human-like visual reading, and takes a step toward more natural and cognitively inspired language models.
  arXiv  Detail & Related papers  (2025-10-21T17:34:48Z)
• False Friends Are Not Foes: Investigating Vocabulary Overlap in Multilingual Language Models [53.01170039144264]
  Subword tokenizers trained on multilingual corpora naturally produce overlapping tokens across languages.<n>Does token overlap facilitate cross-lingual transfer or instead introduce interference between languages?<n>We find that models with overlap outperform models with disjoint vocabularies.
  arXiv  Detail & Related papers  (2025-09-23T07:47:54Z)
• From Tokens to Words: On the Inner Lexicon of LLMs [7.148628740938674]
  Natural language is composed of words, but modern large language models (LLMs) process sub-words as input.<n>We present evidence that LLMs engage in an intrinsic detokenization process, where sub-word sequences are combined into coherent whole-word representations.<n>Our findings suggest that LLMs maintain a latent vocabulary beyond the tokenizer's scope.
  arXiv  Detail & Related papers  (2024-10-08T09:53:35Z)
• From Characters to Words: Hierarchical Pre-trained Language Model for Open-vocabulary Language Understanding [22.390804161191635]
  Current state-of-the-art models for natural language understanding require a preprocessing step to convert raw text into discrete tokens. This process known as tokenization relies on a pre-built vocabulary of words or sub-word morphemes. We introduce a novel open-vocabulary language model that adopts a hierarchical two-level approach.
  arXiv  Detail & Related papers  (2023-05-23T23:22:20Z)
• DP-Parse: Finding Word Boundaries from Raw Speech with an Instance Lexicon [18.05179713472479]
  We introduce DP-Parse, which uses similar principles but only relies on an instance lexicon of word tokens. On the Zero Resource Speech Benchmark 2017, our model sets a new speech segmentation state-of-the-art in 5 languages. Despite lacking a type lexicon, DP-Parse can be pipelined to a language model and learn and semantic representations as assessed by a new spoken word embedding benchmark.
  arXiv  Detail & Related papers  (2022-06-22T19:15:57Z)
• More Than Words: Collocation Tokenization for Latent Dirichlet Allocation Models [71.42030830910227]
  We propose a new metric for measuring the clustering quality in settings where the models differ. We show that topics trained with merged tokens result in topic keys that are clearer, more coherent, and more effective at distinguishing topics than those unmerged models.
  arXiv  Detail & Related papers  (2021-08-24T14:08:19Z)
• Charformer: Fast Character Transformers via Gradient-based Subword Tokenization [50.16128796194463]
  We propose a new model inductive bias that learns a subword tokenization end-to-end as part of the model. We introduce a soft gradient-based subword tokenization module (GBST) that automatically learns latent subword representations from characters. We additionally introduce Charformer, a deep Transformer model that integrates GBST and operates on the byte level.
  arXiv  Detail & Related papers  (2021-06-23T22:24:14Z)
• Char2Subword: Extending the Subword Embedding Space Using Robust Character Compositionality [24.80654159288458]
  We propose a character-based subword module (char2subword) that learns the subword embedding table in pre-trained models like BERT. Our module is robust to character-level alterations such as misspellings, word inflection, casing, and punctuation. We show that incorporating our module to mBERT significantly improves the performance on the social media linguistic code-switching evaluation (LinCE) benchmark.
  arXiv  Detail & Related papers  (2020-10-24T01:08:28Z)
• Grounded Compositional Outputs for Adaptive Language Modeling [59.02706635250856]
  A language model's vocabulary$-$typically selected before training and permanently fixed later$-$affects its size. We propose a fully compositional output embedding layer for language models. To our knowledge, the result is the first word-level language model with a size that does not depend on the training vocabulary.
  arXiv  Detail & Related papers  (2020-09-24T07:21:14Z)
• Supervised Understanding of Word Embeddings [1.160208922584163]
  We have obtained supervised projections in the form of the linear keyword-level classifiers on word embeddings. We have shown that the method creates interpretable projections of original embedding dimensions.
  arXiv  Detail & Related papers  (2020-06-23T20:13:42Z)
• Anchor & Transform: Learning Sparse Embeddings for Large Vocabularies [60.285091454321055]
  We design a simple and efficient embedding algorithm that learns a small set of anchor embeddings and a sparse transformation matrix. On text classification, language modeling, and movie recommendation benchmarks, we show that ANT is particularly suitable for large vocabulary sizes.
  arXiv  Detail & Related papers  (2020-03-18T13:07:51Z)

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.