Saved in:
Bibliographic Details
Main Authors: Xiong, Jing, Chen, Qiujiang, Ye, Fanghua, Wan, Zhongwei, Zheng, Chuanyang, Zhao, Chenyang, Shen, Hui, Li, Hanbo, Tao, Chaofan, Tan, Haochen, Bai, Haoli, Shang, Lifeng, Kong, Lingpeng, Wong, Ngai
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.15148
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918352229761024
author Xiong, Jing
Chen, Qiujiang
Ye, Fanghua
Wan, Zhongwei
Zheng, Chuanyang
Zhao, Chenyang
Shen, Hui
Li, Hanbo
Tao, Chaofan
Tan, Haochen
Bai, Haoli
Shang, Lifeng
Kong, Lingpeng
Wong, Ngai
author_facet Xiong, Jing
Chen, Qiujiang
Ye, Fanghua
Wan, Zhongwei
Zheng, Chuanyang
Zhao, Chenyang
Shen, Hui
Li, Hanbo
Tao, Chaofan
Tan, Haochen
Bai, Haoli
Shang, Lifeng
Kong, Lingpeng
Wong, Ngai
contents Large language models (LLMs) benefit from test-time scaling but are often hampered by high inference latency. Speculative decoding is a natural way to accelerate the scaling process; however, scaling along both the parallel and sequential dimensions poses significant challenges, including substantial memory-bound execution and synchronization overhead. We introduce ATTS (Asynchronous Test-Time Scaling), a statistically guaranteed adaptive scaling framework that follows the hypothesis testing process to address these challenges. By revisiting arithmetic intensity, ATTS identifies synchronization as the primary bottleneck. It enables asynchronous inference through online calibration and proposes an ordinal classification algorithm that supports a three-stage rejection sampling pipeline, scaling along both the sequential and parallel axes. Across experiments on the MATH, AMC23, AIME24, and AIME25 datasets and across multiple draft-target model families, we show that ATTS delivers up to 56.7x speedup in test-time scaling and a 4.14x throughput improvement, while maintaining accurate control of the rejection rate, reducing latency and memory overhead, and incurring no accuracy loss. By scaling both in parallel and sequential dimensions, we enable the 1.5B/70B draft/target model combination to achieve the performance of the state-of-the-art reasoning model o3-mini (high) on the AIME dataset. We have released the code at https://github.com/menik1126/asynchronous-test-time-scaling.
format Preprint
id arxiv_https___arxiv_org_abs_2509_15148
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ATTS: Asynchronous Test-Time Scaling via Conformal Prediction
Xiong, Jing
Chen, Qiujiang
Ye, Fanghua
Wan, Zhongwei
Zheng, Chuanyang
Zhao, Chenyang
Shen, Hui
Li, Hanbo
Tao, Chaofan
Tan, Haochen
Bai, Haoli
Shang, Lifeng
Kong, Lingpeng
Wong, Ngai
Computation and Language
Large language models (LLMs) benefit from test-time scaling but are often hampered by high inference latency. Speculative decoding is a natural way to accelerate the scaling process; however, scaling along both the parallel and sequential dimensions poses significant challenges, including substantial memory-bound execution and synchronization overhead. We introduce ATTS (Asynchronous Test-Time Scaling), a statistically guaranteed adaptive scaling framework that follows the hypothesis testing process to address these challenges. By revisiting arithmetic intensity, ATTS identifies synchronization as the primary bottleneck. It enables asynchronous inference through online calibration and proposes an ordinal classification algorithm that supports a three-stage rejection sampling pipeline, scaling along both the sequential and parallel axes. Across experiments on the MATH, AMC23, AIME24, and AIME25 datasets and across multiple draft-target model families, we show that ATTS delivers up to 56.7x speedup in test-time scaling and a 4.14x throughput improvement, while maintaining accurate control of the rejection rate, reducing latency and memory overhead, and incurring no accuracy loss. By scaling both in parallel and sequential dimensions, we enable the 1.5B/70B draft/target model combination to achieve the performance of the state-of-the-art reasoning model o3-mini (high) on the AIME dataset. We have released the code at https://github.com/menik1126/asynchronous-test-time-scaling.
title ATTS: Asynchronous Test-Time Scaling via Conformal Prediction
topic Computation and Language
url https://arxiv.org/abs/2509.15148