Speculative Decoding for Multi-Sample Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Yiwei, Shi, Jiayi, Feng, Shaoxiong, Yuan, Peiwen, Wang, Xinglin, Zhang, Yueqi, Zhang, Ji, Tan, Chuyi, Pan, Boyuan, Hu, Yao, Li, Kan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910863486615552
author Li, Yiwei
Shi, Jiayi
Feng, Shaoxiong
Yuan, Peiwen
Wang, Xinglin
Zhang, Yueqi
Zhang, Ji
Tan, Chuyi
Pan, Boyuan
Hu, Yao
Li, Kan
author_facet Li, Yiwei
Shi, Jiayi
Feng, Shaoxiong
Yuan, Peiwen
Wang, Xinglin
Zhang, Yueqi
Zhang, Ji
Tan, Chuyi
Pan, Boyuan
Hu, Yao
Li, Kan
contents We propose a novel speculative decoding method tailored for multi-sample reasoning scenarios, such as self-consistency and Best-of-N sampling. Our method exploits the intrinsic consensus of parallel generation paths to synthesize high-quality draft tokens without requiring auxiliary models or external databases. By dynamically analyzing structural patterns across parallel reasoning paths through a probabilistic aggregation mechanism, it identifies consensus token sequences that align with the decoding distribution. Evaluations on mathematical reasoning benchmarks demonstrate a substantial improvement in draft acceptance rates over baselines, while reducing the latency in draft token construction. This work establishes a paradigm shift for efficient multi-sample inference, enabling seamless integration of speculative decoding with sampling-based reasoning techniques.
format Preprint
id arxiv_https___arxiv_org_abs_2503_05330
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Speculative Decoding for Multi-Sample Inference
Li, Yiwei
Shi, Jiayi
Feng, Shaoxiong
Yuan, Peiwen
Wang, Xinglin
Zhang, Yueqi
Zhang, Ji
Tan, Chuyi
Pan, Boyuan
Hu, Yao
Li, Kan
Computation and Language
Artificial Intelligence
We propose a novel speculative decoding method tailored for multi-sample reasoning scenarios, such as self-consistency and Best-of-N sampling. Our method exploits the intrinsic consensus of parallel generation paths to synthesize high-quality draft tokens without requiring auxiliary models or external databases. By dynamically analyzing structural patterns across parallel reasoning paths through a probabilistic aggregation mechanism, it identifies consensus token sequences that align with the decoding distribution. Evaluations on mathematical reasoning benchmarks demonstrate a substantial improvement in draft acceptance rates over baselines, while reducing the latency in draft token construction. This work establishes a paradigm shift for efficient multi-sample inference, enabling seamless integration of speculative decoding with sampling-based reasoning techniques.
title Speculative Decoding for Multi-Sample Inference
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2503.05330