Strategic Scaling of Test-Time Compute: A Bandit Learning Approach

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zuo, Bowen, Zhu, Yinglun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913056773111808
author Zuo, Bowen
Zhu, Yinglun
author_facet Zuo, Bowen
Zhu, Yinglun
contents Scaling test-time compute has emerged as an effective strategy for improving the performance of large language models. However, existing methods typically allocate compute uniformly across all queries, overlooking variation in query difficulty. To address this inefficiency, we formulate test-time compute allocation as a novel bandit learning problem and propose adaptive algorithms that estimate query difficulty on the fly and allocate compute accordingly. Compared to uniform allocation, our algorithms allocate more compute to challenging queries while maintaining accuracy on easier ones. Among challenging queries, our algorithms further learn to prioritize solvable instances, effectively reducing excessive computing on unsolvable queries. We theoretically prove that our algorithms achieve better compute efficiency than uniform allocation and empirically validate their effectiveness on math and code benchmarks. Specifically, our algorithms achieve up to an 11.10% performance improvement (15.04% relative) on the MATH-500 dataset, up to 10.82% (14.44% relative) on the AIME25 dataset, and up to an 11.23% performance improvement (15.29% relative) on the LiveCodeBench dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2506_12721
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Strategic Scaling of Test-Time Compute: A Bandit Learning Approach
Zuo, Bowen
Zhu, Yinglun
Artificial Intelligence
Computation and Language
Machine Learning
Scaling test-time compute has emerged as an effective strategy for improving the performance of large language models. However, existing methods typically allocate compute uniformly across all queries, overlooking variation in query difficulty. To address this inefficiency, we formulate test-time compute allocation as a novel bandit learning problem and propose adaptive algorithms that estimate query difficulty on the fly and allocate compute accordingly. Compared to uniform allocation, our algorithms allocate more compute to challenging queries while maintaining accuracy on easier ones. Among challenging queries, our algorithms further learn to prioritize solvable instances, effectively reducing excessive computing on unsolvable queries. We theoretically prove that our algorithms achieve better compute efficiency than uniform allocation and empirically validate their effectiveness on math and code benchmarks. Specifically, our algorithms achieve up to an 11.10% performance improvement (15.04% relative) on the MATH-500 dataset, up to 10.82% (14.44% relative) on the AIME25 dataset, and up to an 11.23% performance improvement (15.29% relative) on the LiveCodeBench dataset.
title Strategic Scaling of Test-Time Compute: A Bandit Learning Approach
topic Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2506.12721