@inproceedings{syvokon-2026-dictionary,
    title = "Dictionary-Based Speculative Decoding for Non-{L}atin-Script Languages",
    author = "Syvokon, Oleksiy",
    editor = "Romanyshyn, Mariana",
    booktitle = "Proceedings of the Fifth {U}krainian Natural Language Processing Conference ({UNLP} 2026)",
    month = may,
    year = "2026",
    address = "Lviv, Ukraine",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/bulk-corrections-2026-07-02/2026.unlp-1.15/",
    pages = "169--183",
    ISBN = "979-8-89176-359-3",
    abstract = "Large language models tokenize non-Latin-script languagesinefficiently: a single word in Ukrainian or Crimean Tatar is split intotwo to three times as many tokens as its English equivalent. We propose{\_}dictionary-based speculative decoding{\_} (DictSpec), which acceleratesinference by proposing draft continuations from a static n-gram lookuptable built offline from an unlabeled corpus. The lookup table requiresno trainable parameters or GPU resources, is inexpensive to construct,adds under 5 MB of memory overhead, and can be reused across modelsthat share a tokenizer. We evaluate DictSpec on Ukrainian and Crimean Tatar(Cyrillic and Latin scripts), implementing a vLLM plugin to benchmarkfive models ranging from 3B to 70B parameters on consumer- andserver-grade GPUs. In controlled emulation, DictSpec reduces verificationsteps by up to 1.65{\texttimes}, with gains correlating substantially with tokenizerfertility. In live vLLM serving, pure DictSpec gives modest speedups,while a hybrid with prompt-local n-gram speculation reaches up to 1.76{\texttimes}.We release our code and vLLM plugin as opensource."
}