REVIEW 8 cited by
SuperBPE: Space Travel for Language Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
SuperBPE: Space Travel for Language Models
read the original abstract
The assumption across nearly all language model (LM) tokenization schemes is that tokens should be subwords, i.e., contained within word boundaries. While providing a seemingly reasonable inductive bias, is this common practice limiting the potential of modern LMs? Whitespace is not a reliable delimiter of meaning, as evidenced by multi-word expressions (e.g., "by the way"), crosslingual variation in the number of words needed to express a concept (e.g., "spacesuit helmet" in German is "raumanzughelm"), and languages that do not use whitespace at all (e.g., Chinese). To explore the potential of tokenization beyond subwords, we introduce a "superword" tokenizer, SuperBPE, which incorporates a simple pretokenization curriculum into the byte-pair encoding (BPE) algorithm to first learn subwords, then superwords that bridge whitespace. This brings dramatic improvements in encoding efficiency: when fixing the vocabulary size to 200k, SuperBPE encodes a fixed piece of text with up to 33% fewer tokens than BPE on average. In experiments, we pretrain 8B transformer LMs from scratch while fixing the model size, vocabulary size, and train compute, varying *only* the algorithm for learning the vocabulary. Our model trained with SuperBPE achieves an average +4.0% absolute improvement over the BPE baseline across 30 downstream tasks (including +8.2% on MMLU), while simultaneously requiring 27% less compute at inference time. In analysis, we find that SuperBPE results in segmentations of text that are more uniform in per-token difficulty. Qualitatively, this may be because SuperBPE tokens often capture common multi-word expressions that function semantically as a single unit. SuperBPE is a straightforward, local modification to tokenization that improves both encoding efficiency and downstream performance, yielding better language models overall.
Forward citations
Cited by 8 Pith papers
-
BrahmicTokenizer-131K: An Indic-Capable Drop-In Replacement for o200k_base
BrahmicTokenizer-131K is a 131K-vocab tokenizer constructed via script-prune crop and linear-programming retrofit to o200k_base, achieving 26.7% fewer tokens on Indic text while matching o200k_base on English fertilit...
-
Faster Superword Tokenization
Frequency aggregation of supermerge candidates and a two-phase formulation make BoundlessBPE and SuperBPE training over 600x faster on 1GB data while preserving identical results, with open-source Python and Rust code.
-
DC-DiT: Adaptive Compute and Elastic Inference for Visual Generation via Dynamic Chunking
DC-DiT learns dynamic chunking to allocate fewer tokens to smooth or noisy regions and more to detailed or late-stage areas, cutting inference FLOPs up to 36.8% while improving FID up to 37.8% on class-conditional Ima...
-
Vocab Diet: Reshaping the Vocabulary of LLMs via Vector Arithmetic
LLMs can compose surface-form tokens from base embeddings plus learned transformation vectors, freeing 10-40% of vocabulary slots while expanding coverage and preserving downstream performance across five languages.
-
Sampling from Your Language Model One Byte at a Time
An inference-time technique turns BPE-based LMs into byte- or character-level models, solving the prompt boundary problem while unifying vocabularies across different tokenizers.
-
Where to cut, how deep: BPE and Unigram-LM on chemistry SMILES
On chemistry SMILES with a fixed 165-token base, BPE and Unigram-LM produce near-disjoint vocabularies (Jaccard ≤0.161) and Unigram-LM emits 29–41% more tokens across 22 matched conditions.
-
Compute Optimal Tokenization
In compute-optimal regimes, language model parameter count scales proportionally with data bytes rather than tokens, and the optimal compression rate decreases with increasing compute.
-
Compute Optimal Tokenization
Compute-optimal language models require parameter count to scale with data bytes rather than tokens, with optimal token compression rate decreasing as compute budget grows.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.