paper-with-me

Papers

BPE Stays on SCRIPT: Structured Encoding for Robust Multilingual Pretokenization

2025-05-30 · Sander Land, Catherine Arnett

Byte Pair Encoding (BPE) tokenizers, widely used in Large Language Models, face challenges in multilingual settings, including penalization of non-Western scripts and the creation of tokens with partial UTF-8 sequences. Pretokenization, often reliant on complex regular expressions, can also introduce fragility and unexpected edge cases. We propose SCRIPT (Script Category Representation in PreTokenization), a novel encoding scheme that bypasses UTF-8 byte conversion by using initial tokens based on Unicode script and category properties. This approach enables a simple, rule-based pretokenization strategy that respects script boundaries, offering a robust alternative to pretokenization strategies based on regular expressions. We also introduce and validate a constrained BPE merging strategy that enforces character integrity, applicable to both SCRIPT-BPE and byte-based BPE. Our experiments demonstrate that SCRIPT-BPE achieves competitive compression while eliminating encoding-based penalties for non-Latin-script languages.

📄 PDF Abstract BibTeX arXiv:2505.24689

Code (1)

sanderland/script_bpe 공식 구현

Methods 이 논문이 사용한 방법론

BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…

Similar Papers 제목 키워드 기반

Peek2: Regex-free Byte-level Byte-Pair Encoding Pretokenizer for LLM Inference on Edge Devices

2026-01-09 · Liu Zai, Iraklis Klampanos arxiv

Pretokenization is a crucial, sequential pass in Byte-level BPE tokenizers, yet little work has been done to optimize it for edge-side inference. Our proposed new implementation, Peek2, serves as a drop-in replacement fo…

The Art of Breaking Words: Rethinking Multilingual Tokenizer Design

2025-08-03 · Aamod Thakur, Ajay Nagpal, Atharva Savarkar, Kundeshwar Pundalik 외 arxiv

While model architecture and training objectives are well-studied, tokenization, particularly in multilingual contexts, remains a relatively neglected aspect of Large Language Model (LLM) development. Existing tokenizers…

Multilingual Dialogue Generation and Localization with Dialogue Act Scripting

2025-09-26 · Justin Vasselli, Eunike Andriani Kardinata, Yusuke Sakai, Taro Watanabe arxiv

Non-English dialogue datasets are scarce, and models are often trained or evaluated on translations of English-language dialogues, an approach which can introduce artifacts that reduce their naturalness and cultural appr…

Dialogue Generation

MLingualFC: Evaluating Jailbreak Vulnerabilities in Multilingual Vision-Language Models

2026-06-05 · Rishabh Makwana, Mamta, Deeksha Varshney, Oana Cocarascu arxiv

Vision-Language Models (VLMs) have demonstrated strong performance across multimodal tasks, yet their safety robustness remains an open challenge. While prior work has shown that structured visual prompts such as flowcha…

MYTE: Morphology-Driven Byte Encoding for Better and Fairer Multilingual Language Modeling

2024-03-15 · Tomasz Limisiewicz, Terra Blevins, Hila Gonen, Orevaoghene Ahia 외

A major consideration in multilingual language modeling is how to best represent languages with diverse vocabularies and scripts. Although contemporary text encoding methods cover most of the world's writing systems, the…

Language ModelingLanguage Modelling