Reliable Reasoning Beyond Natural Language

TL;DR

This paper introduces the NLR dataset to evaluate reasoning in LLMs and proposes a neurosymbolic approach combining Prolog with LLMs, significantly improving reasoning accuracy and robustness.

Contribution

The paper presents a new dataset for reasoning challenges and a neurosymbolic method that enhances LLM reasoning capabilities beyond natural language limitations.

Findings

Achieves near-perfect accuracy on NLR problems

Significant performance gains on GSM8k and BIG-bench Navigate

Robustness maintained with increasing variable interdependence

Abstract

Despite their linguistic competence, Large Language Models (LLMs) often struggle to reason reliably and flexibly. To identify these shortcomings, we introduce the Non-Linear Reasoning (NLR) dataset, a collection of 55 unique, hand-designed problems that target reasoning bottlenecks arising from the sequential prediction paradigm of LLMs and the inherently linear nature of natural language. NLR tasks require iterative updates, backtracking, and reasoning across multiple parallel chains of thought but only basic arithmetic to solve. To address these limitations, we propose a neurosymbolic reasoning approach that integrates Prolog, a symbolic reasoning engine, into the inference pipeline of LLMs. This division of labor shifts the LLM's task from iterative computations to inferring all information, explicit or implied through common sense, and encoding it as logical code. Our method yields large and robust performance gains across the GSM8k and BIG-bench Navigate benchmarks and achieves near-perfect accuracy on NLR problems, maintaining robustness even as variable interdependence - the number of other variables on which the value of a single variable depends - increases.