<?xml version="1.0" encoding="UTF-8"?>

<modsCollection xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.loc.gov/mods/v3" xsi:schemaLocation="http://www.loc.gov/mods/v3 http://www.loc.gov/standards/mods/v3/mods-3-3.xsd">
<mods version="3.3">

<genre>conference paper</genre>

<titleInfo><title>Speculative decoding speed-of-light: Optimal lower bounds via branching random walks</title></titleInfo>


<note type="publicationStatus">published</note>


<note type="qualityControlled">yes</note>

<name type="personal">
  <namePart type="given">Sergei</namePart>
  <namePart type="family">Pankratov</namePart>
  <role><roleTerm type="text">author</roleTerm> </role><identifier type="local">f773bf05-72ef-11ef-b75a-a383d22f454b</identifier></name>
<name type="personal">
  <namePart type="given">Dan-Adrian</namePart>
  <namePart type="family">Alistarh</namePart>
  <role><roleTerm type="text">author</roleTerm> </role><identifier type="local">4A899BFC-F248-11E8-B48F-1D18A9856A87</identifier><description xsi:type="identifierDefinition" type="orcid">0000-0003-3650-940X</description></name>







<name type="corporate">
  <namePart></namePart>
  <identifier type="local">DaAl</identifier>
  <role>
    <roleTerm type="text">department</roleTerm>
  </role>
</name>

<name type="corporate">
  <namePart></namePart>
  <identifier type="local">GradSch</identifier>
  <role>
    <roleTerm type="text">department</roleTerm>
  </role>
</name>



<name type="conference">
  <namePart>EACL:  Conference of the European Chapter of the Association for Computational Linguistics</namePart>
</name>






<abstract lang="eng">Speculative generation has emerged as a promising technique to accelerate inference in large language models (LLMs) by leveraging parallelism to verify multiple draft tokens simultaneously. However, the fundamental limits on the achievable speedup remain poorly understood. In this work, we establish the first “tight” lower bounds on the runtime of any deterministic speculative generation algorithm. This is achieved by drawing a parallel between the token generation process and branching random walks, which allows us to analyze the optimal draft tree selection problem. We prove, under basic assumptions, that the expected number of tokens successfully predicted per speculative iteration is bounded as \mathbb{E}[X] ≤ (𝜇 + 𝜇(2))log(B )/𝜇2 + O(1), where B is the verifier’s batch size, 𝜇 is the expected entropy of the verifier’s output distribution, and 𝜇(2) is this entropy’s second moment. This result provides new insights into the limits of parallel token generation, and could guide the design of future speculative decoding systems. Empirical evaluations on Llama models validate our theoretical predictions, confirming the tightness of our bounds in practical settings.</abstract>

<originInfo><publisher>Association for Computational Linguistics</publisher><dateIssued encoding="w3cdtf">2026</dateIssued><place><placeTerm type="text">Rabat, Morocco</placeTerm></place>
</originInfo>
<language><languageTerm authority="iso639-2b" type="code">eng</languageTerm>
</language>



<relatedItem type="host"><titleInfo><title>Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics</title></titleInfo>
  <identifier type="arXiv">2512.11718</identifier><identifier type="doi">10.18653/v1/2026.eacl-long.301</identifier>
<part><extent unit="pages">6404–6418</extent>
</part>
</relatedItem>


<extension>
<bibliographicCitation>
<ista>Pankratov S, Alistarh D-A. 2026. Speculative decoding speed-of-light: Optimal lower bounds via branching random walks. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics. EACL:  Conference of the European Chapter of the Association for Computational Linguistics, 6404–6418.</ista>
<chicago>Pankratov, Sergei, and Dan-Adrian Alistarh. “Speculative Decoding Speed-of-Light: Optimal Lower Bounds via Branching Random Walks.” In &lt;i&gt;Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics&lt;/i&gt;, 6404–6418. Association for Computational Linguistics, 2026. &lt;a href=&quot;https://doi.org/10.18653/v1/2026.eacl-long.301&quot;&gt;https://doi.org/10.18653/v1/2026.eacl-long.301&lt;/a&gt;.</chicago>
<apa>Pankratov, S., &amp;#38; Alistarh, D.-A. (2026). Speculative decoding speed-of-light: Optimal lower bounds via branching random walks. In &lt;i&gt;Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics&lt;/i&gt; (pp. 6404–6418). Rabat, Morocco: Association for Computational Linguistics. &lt;a href=&quot;https://doi.org/10.18653/v1/2026.eacl-long.301&quot;&gt;https://doi.org/10.18653/v1/2026.eacl-long.301&lt;/a&gt;</apa>
<ama>Pankratov S, Alistarh D-A. Speculative decoding speed-of-light: Optimal lower bounds via branching random walks. In: &lt;i&gt;Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics&lt;/i&gt;. Association for Computational Linguistics; 2026:6404–6418. doi:&lt;a href=&quot;https://doi.org/10.18653/v1/2026.eacl-long.301&quot;&gt;10.18653/v1/2026.eacl-long.301&lt;/a&gt;</ama>
<ieee>S. Pankratov and D.-A. Alistarh, “Speculative decoding speed-of-light: Optimal lower bounds via branching random walks,” in &lt;i&gt;Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics&lt;/i&gt;, Rabat, Morocco, 2026, pp. 6404–6418.</ieee>
<mla>Pankratov, Sergei, and Dan-Adrian Alistarh. “Speculative Decoding Speed-of-Light: Optimal Lower Bounds via Branching Random Walks.” &lt;i&gt;Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics&lt;/i&gt;, Association for Computational Linguistics, 2026, pp. 6404–6418, doi:&lt;a href=&quot;https://doi.org/10.18653/v1/2026.eacl-long.301&quot;&gt;10.18653/v1/2026.eacl-long.301&lt;/a&gt;.</mla>
<short>S. Pankratov, D.-A. Alistarh, in:, Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics, Association for Computational Linguistics, 2026, pp. 6404–6418.</short>
</bibliographicCitation>
</extension>
<recordInfo><recordIdentifier>22302</recordIdentifier><recordCreationDate encoding="w3cdtf">2026-07-13T10:48:03Z</recordCreationDate><recordChangeDate encoding="w3cdtf">2026-07-14T06:18:11Z</recordChangeDate>
</recordInfo>
</mods>
</modsCollection>
