# Where to cut, how deep: BPE and Unigram-LM on chemistry SMILES

arXiv: https://arxiv.org/abs/2607.05691 · HTML: https://mlsanity.com/paper/2607.05691/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 55 references that are in the map.
- First posted on arXiv 6 Jul 2026.

## Builds on (most cited first, 27 of 27)

- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing](https://mlsanity.com/paper/1808.06226/) (2018)
- [MoleculeNet: A Benchmark for Molecular Machine Learning](https://mlsanity.com/paper/1703.00564/) (2017)
- [Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates](https://mlsanity.com/paper/1804.10959/) (2018)
- [ChemBERTa: Large-Scale Self-Supervised Pretraining for Molecular Property Prediction](https://mlsanity.com/paper/2010.09885/) (2020)
- [How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models](https://mlsanity.com/paper/2012.15613/) (2020)
- [BPE-Dropout: Simple and Effective Subword Regularization](https://mlsanity.com/paper/1910.13267/) (2019)
- [Byte Pair Encoding is Suboptimal for Language Model Pretraining](https://mlsanity.com/paper/2004.03720/) (2020)
- [Tokenizer Choice For LLM Training: Negligible or Crucial?](https://mlsanity.com/paper/2310.08754/) (2023)
- [Tokenization Is More Than Compression](https://mlsanity.com/paper/2402.18376/) (2024)
- [Dynamic Chunking for End-to-End Hierarchical Sequence Modeling](https://mlsanity.com/paper/2507.07955/) (2025)
- [SuperBPE: Space Travel for Language Models](https://mlsanity.com/paper/2503.13423/) (2025)
- [Tokenization and the Noiseless Channel](https://mlsanity.com/paper/2306.16842/) (2023)
- ["Found in Translation": Predicting Outcomes of Complex Organic Chemistry Reactions using Neural Sequence-to-Sequence Models](https://mlsanity.com/paper/1711.04810/) (2017)
- [Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models](https://mlsanity.com/paper/2405.05417/) (2024)
- [Boundless Byte Pair Encoding: Breaking the Pre-tokenization Barrier](https://mlsanity.com/paper/2504.00178/) (2025)
- [Hierarchical Autoregressive Transformers: Combining Byte- and Word-Level Processing for Robust, Adaptable Language Models](https://mlsanity.com/paper/2501.10322/) (2025)
- [BARTSmiles: Generative Masked Language Models for Molecular Representations](https://mlsanity.com/paper/2211.16349/) (2022)
- [Scaffold-BPE: Enhancing Byte Pair Encoding for Large Language Models with Simple and Effective Scaffold Token Removal](https://mlsanity.com/paper/2404.17808/) (2024)
- [How Much is Enough? The Diminishing Returns of Tokenization Training Data](https://mlsanity.com/paper/2502.20273/) (2025)
- [ReactionT5: a large-scale pre-trained model towards application of limited reaction data](https://mlsanity.com/paper/2311.06708/) (2023)
- [Two Counterexamples to Tokenization and the Noiseless Channel](https://mlsanity.com/paper/2402.14614/) (2024)
- [Tokenization is Sensitive to Language Variation](https://mlsanity.com/paper/2502.15343/) (2025)
- [Linguistic Laws Meet Protein Sequences: A Comparative Analysis of Subword Tokenization Methods](https://mlsanity.com/paper/2411.17669/) (2024)
- [Stop Taking Tokenizers for Granted: They Are Core Design Decisions in Large Language Models](https://mlsanity.com/paper/2601.13260/) (2026)
- [Tokenization for Molecular Foundation Models](https://mlsanity.com/paper/2409.15370/) (2024)
- [The Tokenization Bottleneck: How Vocabulary Extension Improves Chemistry Representation Learning in Pretrained Language Models](https://mlsanity.com/paper/2511.14365/) (2025)

## Abstract (arXiv)

Every chemical language model reading SMILES begins with a tokenizer, yet the field has inherited byte-pair encoding (BPE) from natural language with little scrutiny. In natural language, BPE's principal alternative, Unigram-LM, is known to build structurally different vocabularies. Whether that contrast survives in chemistry was open. We report a controlled comparison of BPE and Unigram-LM over a fixed 165-token chemistry base, at the small vocabulary sizes where token embeddings are learnable, across three corpus typologies (diverse, drug-like, natural-products) and both pre-tokenization boundary policies. The two do not converge. In all 22 matched conditions they build near-disjoint subword vocabularies: cross-algorithm Jaccard overlap on the learned pieces never exceeds 0.161, and at most 0.05 once weighted toward the high-frequency pieces a model updates most. Unigram-LM also segments held-out molecules into 29-41% more tokens; the arms largely agree on where to cut but not how deeply, so BPE's segmentation is a strict coarsening of Unigram-LM's on 80-99% of molecules. The separation holds across corpus, boundary, and vocabulary size, persisting even at eight times that scale. The subword algorithm is therefore a modeling decision, not a free default. The study trains no language models.