@inproceedings{kobus-gunduz-2025-speculative,
title = "Speculative Sampling via Exponential Races",
author = "Kobus, Szymon and
Gunduz, Deniz",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.936/",
doi = "10.18653/v1/2025.findings-acl.936",
pages = "18189--18204",
ISBN = "979-8-89176-256-5",
abstract = "Speculative decoding accelerates large language model inference using a smaller draft model. In this paper, we establish a surprising connection between speculative sampling and the concept of channel simulation from information theory, which aims at simulating a noisy channel using as few bits as possible. This connection allows us to provide an information-theoretic analysis of the speed up that can be achieved by speculative sampling. Leveraging this link, we derive an explicit relation between generation speed-up and the number of tokens $k$ generated by the draft model for large $k$, which serves as an upper bound for all $k$. We also propose a novel speculative sampling method via exponential races called ERSS that matches state-of-the-art performance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kobus-gunduz-2025-speculative">
<titleInfo>
<title>Speculative Sampling via Exponential Races</title>
</titleInfo>
<name type="personal">
<namePart type="given">Szymon</namePart>
<namePart type="family">Kobus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Deniz</namePart>
<namePart type="family">Gunduz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>Speculative decoding accelerates large language model inference using a smaller draft model. In this paper, we establish a surprising connection between speculative sampling and the concept of channel simulation from information theory, which aims at simulating a noisy channel using as few bits as possible. This connection allows us to provide an information-theoretic analysis of the speed up that can be achieved by speculative sampling. Leveraging this link, we derive an explicit relation between generation speed-up and the number of tokens k generated by the draft model for large k, which serves as an upper bound for all k. We also propose a novel speculative sampling method via exponential races called ERSS that matches state-of-the-art performance.</abstract>
<identifier type="citekey">kobus-gunduz-2025-speculative</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.936</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.936/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>18189</start>
<end>18204</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Speculative Sampling via Exponential Races
%A Kobus, Szymon
%A Gunduz, Deniz
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F kobus-gunduz-2025-speculative
%X Speculative decoding accelerates large language model inference using a smaller draft model. In this paper, we establish a surprising connection between speculative sampling and the concept of channel simulation from information theory, which aims at simulating a noisy channel using as few bits as possible. This connection allows us to provide an information-theoretic analysis of the speed up that can be achieved by speculative sampling. Leveraging this link, we derive an explicit relation between generation speed-up and the number of tokens k generated by the draft model for large k, which serves as an upper bound for all k. We also propose a novel speculative sampling method via exponential races called ERSS that matches state-of-the-art performance.
%R 10.18653/v1/2025.findings-acl.936
%U https://aclanthology.org/2025.findings-acl.936/
%U https://doi.org/10.18653/v1/2025.findings-acl.936
%P 18189-18204
Markdown (Informal)
[Speculative Sampling via Exponential Races](https://aclanthology.org/2025.findings-acl.936/) (Kobus & Gunduz, Findings 2025)
ACL
- Szymon Kobus and Deniz Gunduz. 2025. Speculative Sampling via Exponential Races. In Findings of the Association for Computational Linguistics: ACL 2025, pages 18189–18204, Vienna, Austria. Association for Computational Linguistics.