Skip to content
AI News HubLIVE
Original source2 min read

Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding

Summary

AdaptiveSpec is a training-free, per-step speculative decoding method that adapts both token-acceptance margin and draft-tree shape from internal decoding signals. Implemented on SGLang, it improves throughput over EAGLE-3 by up to 56% while recovering 93% to fully lossless task accuracy on GSM8K, MATH-500, and HumanEval.

SourcearXiv Computational LinguisticsAuthor: Oszk\'ar Urb\'an, Young D. Kwon, Stylianos I. Venieris, Cecilia Mascolo
Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 3 Jul 2026]

Title:Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding

View a PDF of the paper titled Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding, by Oszk\'ar Urb\'an and 3 other authors

View PDF HTML (experimental)

Abstract:Speculative decoding accelerates LLM inference by drafting candidate tokens and verifying them in parallel. Tree-attention drafters such as EAGLE-3 are widely adopted, yet typically hold two decisions fixed: (1) a strict token-match verification rule and (2) a static draft-tree shape. Prior work relaxes each in isolation under limiting assumptions: long draft chains for training-free lossy verification, and adaptive tree shaping under a fixed token budget. We introduce AdaptiveSpec, a training-free per-step speculative decoding method that adapts both decisions from internal signals already produced during decoding. A per-step margin rule promotes a mismatched draft-proposed token when the ratio of the target's probability on the drafted token to its top-1 probability exceeds a threshold with no dependence on draft length or underlying drafter architecture. A per-step tree policy adjusts the draft tree's depth, width, and node count directly from a fused signal of draft top-1 confidence and a rolling acceptance history capturing recent draft-target agreement, allowing the total draft count to vary rather than only be redistributed. The two adaptations operate on orthogonal axes and compound in effect. Implemented on the SGLang production-grade serving engine, AdaptiveSpec improves throughput over the state-of-the-art autoregressive speculative decoding method EAGLE-3 by up to 56%, recovering 93% to fully lossless task accuracy across GSM8K, MATH-500, and HumanEval on three target models (DeepSeek-R1-Distill-Llama-8B, Llama-3.1-8B-Instruct, Qwen3-8B).

Subjects:

Computation and Language (cs.CL)

Cite as: arXiv:2609.02897 [cs.CL]

(or arXiv:2609.02897v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2609.02897

arXiv-issued DOI via DataCite

Submission history

From: Oszkár Urbán [view email] [v1] Fri, 3 Jul 2026 20:43:49 UTC (378 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding, by Oszk\'ar Urb\'an and 3 other authors

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.CL

new | recent | 2026-09

Change to browse by:

cs

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • AdaptiveSpec adjusts two previously fixed decisions each step: the lossy token-acceptance rule and the draft-tree structure, without extra training.
  • Its margin rule is independent of draft length and drafter architecture, while its tree policy lets the total draft count vary based on confidence and acceptance history.
  • On SGLang, AdaptiveSpec boosts throughput by up to 56% versus EAGLE-3 while preserving near-lossless to fully lossless accuracy across three 8B models and three benchmarks.

Highlights and analysis are generated automatically and may contain errors. Check the original source.