# Understanding Subword Compositionality of Large Language Models

arXiv: https://arxiv.org/abs/2508.17953 · HTML: https://mlsanity.com/paper/2508.17953/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 42 references that are in the map.
- First posted on arXiv 25 Aug 2025.

## Builds on (most cited first, 28 of 28)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates](https://mlsanity.com/paper/1804.10959/) (2018)
- [The Falcon Series of Open Language Models](https://mlsanity.com/paper/2311.16867/) (2023)
- [Faith and Fate: Limits of Transformers on Compositionality](https://mlsanity.com/paper/2305.18654/) (2023)
- [StarCoder 2 and The Stack v2: The Next Generation](https://mlsanity.com/paper/2402.19173/) (2024)
- [What you can cram into a single vector: Probing sentence embeddings for linguistic properties](https://mlsanity.com/paper/1805.01070/) (2018)
- [Aya Expanse: Combining Research Breakthroughs for a New Multilingual Frontier](https://mlsanity.com/paper/2412.04261/) (2024)
- [Neural Machine Translation with Byte-Level Subwords](https://mlsanity.com/paper/1909.03341/) (2019)
- [Understanding and Patching Compositional Reasoning in LLMs](https://mlsanity.com/paper/2402.14328/) (2024)
- [Fast Vocabulary Transfer for Language Model Compression](https://mlsanity.com/paper/2402.09977/) (2024)
- [Assessing Phrasal Representation and Composition in Transformers](https://mlsanity.com/paper/2010.03763/) (2020)
- [Evaluating Compositionality in Sentence Embeddings](https://mlsanity.com/paper/1802.04302/) (2018)
- [Tokenization Matters! Degrading Large Language Models through Challenging Their Tokenization](https://mlsanity.com/paper/2405.17067/) (2024)
- [Spying on your neighbors: Fine-grained probing of contextual embeddings for information about surrounding words](https://mlsanity.com/paper/2005.01810/) (2020)
- [Can LLMs Reason with Rules? Logic Scaffolding for Stress-Testing and Improving LLMs](https://mlsanity.com/paper/2402.11442/) (2024)
- [Do Vision and Language Models Share Concepts? A Vector Space Alignment Study](https://mlsanity.com/paper/2302.06555/) (2023)
- [Assessing Composition in Sentence Vector Representations](https://mlsanity.com/paper/1809.03992/) (2018)
- [The SIGMORPHON 2022 Shared Task on Morpheme Segmentation](https://mlsanity.com/paper/2206.07615/) (2022)
- [Unnatural Error Correction: GPT-4 Can Almost Perfectly Handle Unnatural Scrambled Text](https://mlsanity.com/paper/2311.18805/) (2023)
- [PIXAR: Auto-Regressive Language Modeling in Pixel Space](https://mlsanity.com/paper/2401.03321/) (2024)
- [Concept Space Alignment in Multilingual LLMs](https://mlsanity.com/paper/2410.01079/) (2024)
- [Autoregressive Pre-Training on Pixels and Texts](https://mlsanity.com/paper/2404.10710/) (2024)
- [On the Tip of the Tongue: Analyzing Conceptual Representation in Large Language Models with Reverse-Dictionary Probe](https://mlsanity.com/paper/2402.14404/) (2024)
- [Geometry of Compositionality](https://mlsanity.com/paper/1611.09799/) (2016)
- [Representing Syntax and Composition with Geometric Transformations](https://mlsanity.com/paper/2106.01904/) (2021)

## Built on by (most cited first, 3 of 3)

- [See the Text: From Tokenization to Visual Reading](https://mlsanity.com/paper/2510.18840/) (2025)
- [Bolmo: Byteifying the Next Generation of Language Models](https://mlsanity.com/paper/2512.15586/) (2025)
- [Evaluating Adjective-Noun Compositionality in LLMs: Functional vs Representational Perspectives](https://mlsanity.com/paper/2603.09994/) (2026)

## Abstract (arXiv)

Large language models (LLMs) take sequences of subwords as input, requiring them to effective compose subword representations into meaningful word-level representations. In this paper, we present a comprehensive set of experiments to probe how LLMs compose subword information, focusing on three key aspects: structural similarity, semantic decomposability, and form retention. Our analysis of the experiments suggests that these five LLM families can be classified into three distinct groups, likely reflecting difference in their underlying composition strategies. Specifically, we observe (i) three distinct patterns in the evolution of structural similarity between subword compositions and whole-word representations across layers; (ii) great performance when probing layer by layer their sensitivity to semantic decompositionality; and (iii) three distinct patterns when probing sensitivity to formal features, e.g., character sequence length. These findings provide valuable insights into the compositional dynamics of LLMs and highlight different compositional pattens in how LLMs encode and integrate subword information.