Long context inference scenarios have become increasingly important for large language models, yet they introduce significant computational latency. While prior research has optimized long-sequence inference through operators, model architectures, and system frameworks, tokenization remains an overlooked bottleneck. Existing parallel tokenization methods accelerate processing through text segmentation and multi-process tokenization, but they suffer from inconsistent results due to boundary artifacts that occur after merging. To address this, we propose LoPT, a novel Lossless Parallel Tokenization framework that ensures output identical to standard sequential tokenization. Our approach employs character-position-based matching and dynamic chunk length adjustment to align and merge tokenized segments accurately. Extensive experiments across diverse long-text datasets demonstrate that LoPT achieves significant speedup while guaranteeing lossless tokenization. We also provide theoretical proof of consistency and comprehensive analytical studies to validate the robustness of our method.
@article{arxiv.2511.04952,
title = {LoPT: Lossless Parallel Tokenization Acceleration for Long Context Inference of Large Language Model},
author = {Wei Shao and Lingchao Zheng and Pengyu Wang and Peizhen Zheng and Jun Li and Yuwei Fan},
journal= {arXiv preprint arXiv:2511.04952},
year = {2025}
}