Saved in:
Bibliographic Details
Main Authors: Wang, Zi, Weng, Shiwei, Alhanahnah, Mohannad, Jha, Somesh, Reps, Tom
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.10938
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915154141118464
author Wang, Zi
Weng, Shiwei
Alhanahnah, Mohannad
Jha, Somesh
Reps, Tom
author_facet Wang, Zi
Weng, Shiwei
Alhanahnah, Mohannad
Jha, Somesh
Reps, Tom
contents Large Language Models (LLMs) have exhibited remarkable capabilities across diverse domains, prompting investigations into their potential as generic reasoning engines. While recent studies have explored inference-time computation to enhance model performance on complex problems, current research lacks a formal framework to characterize the complexity of reasoning tasks. This study introduces the Predicate-Enumeration-Aggregation (PEA) framework, a formal approach to describe and solve a class of important reasoning tasks termed computational reasoning problems. The PEA framework decomposes these problems into predicate and enumeration components, using LLMs to synthesize programs based on specified predicates, enumeration, and aggregation rules. These synthesized programs are then executed to obtain solutions to the computational tasks. We demonstrate the framework's efficacy on benchmark tasks including Boolean satisfiability problems, game of $24$, and planning problems. Empirical evaluation reveals that PEA substantially enhances the performance of underlying models on benchmark computational problems, yielding an average accuracy improvement of approximately $50\%$, coupled with increased efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2502_10938
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PEA: Enhancing LLM Performance on Computational-Reasoning Tasks
Wang, Zi
Weng, Shiwei
Alhanahnah, Mohannad
Jha, Somesh
Reps, Tom
Artificial Intelligence
Large Language Models (LLMs) have exhibited remarkable capabilities across diverse domains, prompting investigations into their potential as generic reasoning engines. While recent studies have explored inference-time computation to enhance model performance on complex problems, current research lacks a formal framework to characterize the complexity of reasoning tasks. This study introduces the Predicate-Enumeration-Aggregation (PEA) framework, a formal approach to describe and solve a class of important reasoning tasks termed computational reasoning problems. The PEA framework decomposes these problems into predicate and enumeration components, using LLMs to synthesize programs based on specified predicates, enumeration, and aggregation rules. These synthesized programs are then executed to obtain solutions to the computational tasks. We demonstrate the framework's efficacy on benchmark tasks including Boolean satisfiability problems, game of $24$, and planning problems. Empirical evaluation reveals that PEA substantially enhances the performance of underlying models on benchmark computational problems, yielding an average accuracy improvement of approximately $50\%$, coupled with increased efficiency.
title PEA: Enhancing LLM Performance on Computational-Reasoning Tasks
topic Artificial Intelligence
url https://arxiv.org/abs/2502.10938