EUVIMEDEUVIMED entdecken
Crossref

AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification

Dinh-Truong Do, Nguyen-Khang Le, Le-Minh Nguyen

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (36) · 30530-30538

DOI: 10.1609/aaai.v40i36.40307 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Dinh-Truong Do und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Speculative decoding accelerates large language model (LLM) inference by using a lightweight drafter to propose multiple tokens, which are then verified in parallel by the base model. While effective in English, existing methods often struggle in multilingual scenarios due to static vocabularies and the lack of language-specific instruction data. To address these limitations, we present AdaSpec, a multilingual speculative decoding framework that dynamically adapts both the drafter and vocabulary at decoding time. AdaSpec generates language-specific instruction data using the LLM itself, enabling training of drafters for low-resource languages. It also constructs adaptive vocabularies tailored to each language's characteristics. In addition, we introduce Multi-SpecBench, a comprehensive multilingual benchmark covering seven languages and seven generation tasks, to evaluate multilingual speculative decoding performance. Extensive experiments show that AdaSpec achieves up to 2.3× speedup over the state-of-the-art method of EAGLE-2, even in English, demonstrating its effectiveness across diverse languages and tasks.

Zitieren

APA

Do, D., Le, N., & Nguyen, L. (2026). AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification. Proceedings of the AAAI Conference on Artificial Intelligence, 40(36), 30530-30538. https://doi.org/10.1609/aaai.v40i36.40307

Vancouver

Do D, Le N, Nguyen L. AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(36):30530-30538. doi:10.1609/aaai.v40i36.40307

BibTeX
@article{Do2026,
  author = {Do, Dinh-Truong and Le, Nguyen-Khang and Nguyen, Le-Minh},
  title = {AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {36},
  pages = {30530--30538},
  doi = {10.1609/aaai.v40i36.40307}
}
Kostenlos starten →