arXiv:2608. 15454v1 Announce Type: new Abstract: Byte-level hierarchical language models (LMs) have recently emerged as a robust alternative to their popular counterparts that use subword tokenization.
Byte-level hierarchical language models (LMs) have recently emerged as a robust alternative to their popular counterparts that use subword tokenization. However, generating one byte at a time remains...
The paper investigates why current byte‑level language models, which use a hierarchical structure that down‑samples to words and then upsamples back to bytes, struggle with precise character manipulation. Experiments show that pure byte‑level models outperform hierarchical variants on character‑level tasks, and that byte‑level attention is the key component driving this advantage. The study explains the trade‑off between computational efficiency and fine‑grained character understanding in hierarchical byte models.
The paper proposes a modular tokenizer framework for multilingual large language models, allowing the creation of language‑specific subtokenizers that match monolingual compression quality. It introduces a pretraining strategy that samples these subtokenizers to limit predictions to relevant vocabularies, enabling efficient training and inference. This approach reduces memory usage and speeds up inference without compromising performance.
Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead to uneven compression across them. Moreover, their large embedding...
Doc2FRC introduces Fixed-Range Chunking (FRC), a dynamic programming method that partitions documents into chunks of a predefined length interval, ensuring consistent length distributions during training and inference. This approach reduces train-test length mismatch, mitigates n-gram repetition, and improves translation quality for 7B LLMs compared to direct Doc2Doc fine-tuning. Experiments on IWSLT2017 and a new 10-language test set, GlobVDoc, demonstrate that FRC outperforms existing document-level machine translation methods and enhances out-of-distribution translation performance.
The paper introduces SALT, a lightweight post‑training technique that injects span‑level supervision into existing cross‑lingual sentence encoders to enhance token representations. Evaluated on five multilingual token‑level benchmarks, SALT achieves the best overall results on four tasks, surpassing alternative fine‑tuning methods and competitive encoders. Additionally, SALT improves sentence‑level performance on cross‑lingual retrieval and classification tasks, demonstrating the effectiveness of span‑level supervision for both token and sentence representations.
This work examines recent byte-level models and their failure to perfectly manipulate characters. State-of-the-art byte-level models use a hierarchical structure, starting at the byte level, downsampl...
The paper introduces TokenAdapt, a model‑agnostic tokenizer transplantation method that uses a hybrid heuristic to initialize new token embeddings, and a novel pre‑tokenization learning approach for multi‑word Supertokens to improve compression. TokenAdapt combines local subword decomposition and global semantic similarity to preserve semantics while reducing retraining needs. Empirical results show that TokenAdapt outperforms existing baselines such as Transtokenizer and ReTok, achieving lower perplexity ratios and significant compression gains.
arXiv:2608. 07629v1 Announce Type: cross Abstract: Multilingual neural machine translation models such as NLLB-200 cover 200 languages but leave thousands unsupported, including most Grassfields Bantu languages of Cameroon.
arXiv:2609.01151v1 Announce Type: new Abstract: In the standard LM training pipeline, subword tokenisation is applied as a preprocessing step. Subword segmental language modelling is an alternative p...
StructLens is a framework that analyzes language model representations by constructing maximum spanning trees from semantic representations in residual streams, inspired by dependency parsing. It reveals that contiguous tokens are close in representation space, with middle layers showing the strongest local-span organization. The study also shows that smaller local units emerge earlier during pre‑training, while larger units appear later, offering insights into how token representations are organized across layers and training stages.