GenSelect: A Generative Approach to Best-of-N

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Toshniwal, Shubham, Sorokin, Ivan, Ficek, Aleksander, Moshkov, Ivan, Gitman, Igor
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909703102005248
author Toshniwal, Shubham
Sorokin, Ivan
Ficek, Aleksander
Moshkov, Ivan
Gitman, Igor
author_facet Toshniwal, Shubham
Sorokin, Ivan
Ficek, Aleksander
Moshkov, Ivan
Gitman, Igor
contents Generative reward models with parallel sampling have enabled effective test-time scaling for reasoning tasks. Current approaches employ pointwise scoring of individual solutions or pairwise comparisons. However, pointwise methods underutilize LLMs' comparative abilities, while pairwise methods scale inefficiently with larger sampling budgets. We introduce GenSelect, where the LLM uses long reasoning to select the best solution among N candidates. This leverages LLMs' comparative strengths while scaling efficiently across parallel sampling budgets. For math reasoning, we demonstrate that reasoning models, such as QwQ and DeepSeek-R1-0528, excel at GenSelect, outperforming existing scoring approaches with simple prompting.
format Preprint
id arxiv_https___arxiv_org_abs_2507_17797
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GenSelect: A Generative Approach to Best-of-N
Toshniwal, Shubham
Sorokin, Ivan
Ficek, Aleksander
Moshkov, Ivan
Gitman, Igor
Machine Learning
Computation and Language
Generative reward models with parallel sampling have enabled effective test-time scaling for reasoning tasks. Current approaches employ pointwise scoring of individual solutions or pairwise comparisons. However, pointwise methods underutilize LLMs' comparative abilities, while pairwise methods scale inefficiently with larger sampling budgets. We introduce GenSelect, where the LLM uses long reasoning to select the best solution among N candidates. This leverages LLMs' comparative strengths while scaling efficiently across parallel sampling budgets. For math reasoning, we demonstrate that reasoning models, such as QwQ and DeepSeek-R1-0528, excel at GenSelect, outperforming existing scoring approaches with simple prompting.
title GenSelect: A Generative Approach to Best-of-N
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2507.17797