Pith. sign in

REVIEW 5 cited by

Charformer: Fast Character Transformers via Gradient-based Subword Tokenization

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

arxiv 2106.12672 v3 pith:CUPLFF23 submitted 2021-06-23 cs.CL cs.AIcs.LG

classification cs.CLcs.AIcs.LG
keywords subwordcharformertokenizationgbstlearnsmodelmodelsadditionally
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

State-of-the-art models in natural language processing rely on separate rigid subword tokenization algorithms, which limit their generalization ability and adaptation to new settings. In this paper, we propose a new model inductive bias that learns a subword tokenization end-to-end as part of the model. To this end, we introduce a soft gradient-based subword tokenization module (GBST) that automatically learns latent subword representations from characters in a data-driven fashion. Concretely, GBST enumerates candidate subword blocks and learns to score them in a position-wise fashion using a block scoring network. We additionally introduce Charformer, a deep Transformer model that integrates GBST and operates on the byte level. Via extensive experiments on English GLUE, multilingual, and noisy text datasets, we show that Charformer outperforms a series of competitive byte-level baselines while generally performing on par and sometimes outperforming subword-based models. Additionally, Charformer is fast, improving the speed of both vanilla byte-level and subword-level Transformers by 28%-100% while maintaining competitive quality. We believe this work paves the way for highly performant token-free models that are trained completely end-to-end.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dynamic Chunking for End-to-End Hierarchical Sequence Modeling

    cs.LG 2025-07 conditional novelty 7.0 of 10

    A learned dynamic chunking hierarchy lets byte-level language models match or beat BPE-tokenized Transformers at matched compute, with larger gains on Chinese, code, and DNA.

  2. Do LLM Modules Generalize? A Study on Motion Generation for Autonomous Driving

    cs.AI 2025-09 conditional novelty 6.0 of 10

    On Waymo Sim Agents, LLM-style tokenization, positional embeddings, pretraining, RL post-training, and test-time search can be adapted to improve motion generation, but not all transfer without domain-specific changes.

  3. chDzDT: Word-level morphology-aware language model for Algerian social media text

    cs.CL 2025-09 conditional novelty 5.0 of 10

    A character-level transformer trained on isolated words, with a five-language label head, yields compact embeddings that rival or beat much larger pre-trained models on morphological tagging across Arabic, English, an...

  4. Return of the Encoder: Maximizing Parameter Efficiency for SLMs

    cs.CL 2025-01 conditional novelty 4.0 of 10

    At equal small sizes, encoder-decoder language models are faster on edge hardware and, after distillation from a larger decoder-only teacher, match or beat decoder-only models on several benchmarks.

  5. Digital Guardians: Can GPT-4, Perspective API, and Moderation API reliably detect hate speech in reader comments of German online newspapers?

    cs.CL 2025-01 conditional novelty 4.0 of 10

    On the German HOCON34k hate speech test set, GPT-4o with one-shot prompting scored highest on the combined F2/MCC metric, about 5 points above the fine-tuned BERT baseline.

Pith tools