Enregistré dans:
Détails bibliographiques
Auteurs principaux: Lall, Supriya, Farrell, Christian, Pathanjaly, Hari, Pavic, Marko, Chezhian, Sarvesh, Asai, Masataro
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2509.08150
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911709572104192
author Lall, Supriya
Farrell, Christian
Pathanjaly, Hari
Pavic, Marko
Chezhian, Sarvesh
Asai, Masataro
author_facet Lall, Supriya
Farrell, Christian
Pathanjaly, Hari
Pavic, Marko
Chezhian, Sarvesh
Asai, Masataro
contents Reasoning is a fundamentally algorithmic task. Yet current work on LLM-based reasoning relies on free-form generation whose theoretical guarantees (soundness, completeness, complexity, optimality) remain poorly understood. We argue that we should not treat them as general-purpose reasoners, and as an alternative, we propose a paradigm we call \emph{verbalized algorithms} (VAs), which combines LLMs and various algorithms with established guarantees. Instead of betting on LLM's ability to solve a reasoning task, VAs limit their scope by decomposing the task down to simple elementary operations on strings that they can answer reliably. For example, sorting a list of natural language strings could be done by using an LLM as a binary comparison oracle in a parallel or approximate sorting algorithm. We push the accuracy-runtime Pareto front with \emph{verbalized maximum}, \emph{sorting}, \emph{clustering}, and \emph{submodular maximization}, for numerical reasoning, topic clustering, Wi-Fi access point optimization, and multi-hop Q\&A RAG task. These results suggest improving LLM-based reasoning through standard algorithmic analysis is a feasible and better grounded research direction.
format Preprint
id arxiv_https___arxiv_org_abs_2509_08150
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Verbalized Algorithms: Classical Algorithms are All You Need (Mostly)
Lall, Supriya
Farrell, Christian
Pathanjaly, Hari
Pavic, Marko
Chezhian, Sarvesh
Asai, Masataro
Computation and Language
Reasoning is a fundamentally algorithmic task. Yet current work on LLM-based reasoning relies on free-form generation whose theoretical guarantees (soundness, completeness, complexity, optimality) remain poorly understood. We argue that we should not treat them as general-purpose reasoners, and as an alternative, we propose a paradigm we call \emph{verbalized algorithms} (VAs), which combines LLMs and various algorithms with established guarantees. Instead of betting on LLM's ability to solve a reasoning task, VAs limit their scope by decomposing the task down to simple elementary operations on strings that they can answer reliably. For example, sorting a list of natural language strings could be done by using an LLM as a binary comparison oracle in a parallel or approximate sorting algorithm. We push the accuracy-runtime Pareto front with \emph{verbalized maximum}, \emph{sorting}, \emph{clustering}, and \emph{submodular maximization}, for numerical reasoning, topic clustering, Wi-Fi access point optimization, and multi-hop Q\&A RAG task. These results suggest improving LLM-based reasoning through standard algorithmic analysis is a feasible and better grounded research direction.
title Verbalized Algorithms: Classical Algorithms are All You Need (Mostly)
topic Computation and Language
url https://arxiv.org/abs/2509.08150