Refining Answer Distributions for Improved Large Language Model Reasoning

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Pal, Soumyasundar, Chételat, Didier, Zhang, Yingxue, Coates, Mark
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913787062255616
author Pal, Soumyasundar
Chételat, Didier
Zhang, Yingxue
Coates, Mark
author_facet Pal, Soumyasundar
Chételat, Didier
Zhang, Yingxue
Coates, Mark
contents Large Language Models (LLMs) have exhibited an impressive capability to perform reasoning tasks, especially if they are encouraged to generate a sequence of intermediate steps. Reasoning performance can be improved by suitably combining multiple LLM responses, generated either in parallel in a single query, or via sequential interactions with LLMs throughout the reasoning process. Existing strategies for combination, such as self-consistency and progressive-hint-prompting, make inefficient usage of the LLM responses. We present Refined Answer Distributions, a novel and principled algorithmic framework to enhance the reasoning capabilities of LLMs. Our approach can be viewed as an iterative sampling strategy for forming a Monte Carlo approximation of an underlying distribution of answers, with the goal of identifying the mode -- the most likely answer. Empirical evaluation on several reasoning benchmarks demonstrates the superiority of the proposed approach.
format Preprint
id arxiv_https___arxiv_org_abs_2412_13292
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Refining Answer Distributions for Improved Large Language Model Reasoning
Pal, Soumyasundar
Chételat, Didier
Zhang, Yingxue
Coates, Mark
Computation and Language
Large Language Models (LLMs) have exhibited an impressive capability to perform reasoning tasks, especially if they are encouraged to generate a sequence of intermediate steps. Reasoning performance can be improved by suitably combining multiple LLM responses, generated either in parallel in a single query, or via sequential interactions with LLMs throughout the reasoning process. Existing strategies for combination, such as self-consistency and progressive-hint-prompting, make inefficient usage of the LLM responses. We present Refined Answer Distributions, a novel and principled algorithmic framework to enhance the reasoning capabilities of LLMs. Our approach can be viewed as an iterative sampling strategy for forming a Monte Carlo approximation of an underlying distribution of answers, with the goal of identifying the mode -- the most likely answer. Empirical evaluation on several reasoning benchmarks demonstrates the superiority of the proposed approach.
title Refining Answer Distributions for Improved Large Language Model Reasoning
topic Computation and Language
url https://arxiv.org/abs/2412.13292