Benchmarking GPT-4 on Algorithmic Problems: A Systematic Evaluation of Prompting Strategies

TL;DR

This paper systematically evaluates GPT-4's performance on algorithmic problems, comparing prompting strategies and models, revealing GPT-4's strong baseline capabilities in challenging tasks requiring systematic generalization.

Contribution

It provides a comprehensive benchmarking of GPT-4 on algorithmic tasks and assesses the impact of advanced prompting techniques on performance.

Findings

GPT-4 outperforms GPT-3.5 and Neural Data Router with advanced prompting.

Prompting strategies significantly improve GPT-4 accuracy.

GPT-4 demonstrates strong systematic generalization in algorithmic tasks.

Abstract

Large Language Models (LLMs) have revolutionized the field of Natural Language Processing thanks to their ability to reuse knowledge acquired on massive text corpora on a wide variety of downstream tasks, with minimal (if any) tuning steps. At the same time, it has been repeatedly shown that LLMs lack systematic generalization, which allows to extrapolate the learned statistical regularities outside the training distribution. In this work, we offer a systematic benchmarking of GPT-4, one of the most advanced LLMs available, on three algorithmic tasks characterized by the possibility to control the problem difficulty with two parameters. We compare the performance of GPT-4 with that of its predecessor (GPT-3.5) and with a variant of the Transformer-Encoder architecture recently introduced to solve similar tasks, the Neural Data Router. We find that the deployment of advanced prompting techniques allows GPT-4 to reach superior accuracy on all tasks, demonstrating that state-of-the-art LLMs constitute a very strong baseline also in challenging tasks that require systematic generalization.