Unnatural Error Correction: GPT-4 Can Almost Perfectly Handle Unnatural Scrambled Text

TL;DR

This paper demonstrates that GPT-4 exhibits near-perfect ability to correct and understand extensively scrambled text, revealing remarkable resilience of LLMs to unnatural input errors, unlike other models.

Contribution

The study introduces the Scrambled Bench to evaluate LLMs' handling of scrambled input and shows GPT-4's exceptional error correction capabilities under extreme conditions.

Findings

GPT-4 reduces edit distance by 95% on scrambled sentences

Most LLMs perform poorly on scrambled input, except GPT-4

GPT-4 nearly flawlessly reconstructs original sentences from scrambled text

Abstract

While Large Language Models (LLMs) have achieved remarkable performance in many tasks, much about their inner workings remains unclear. In this study, we present novel experimental insights into the resilience of LLMs, particularly GPT-4, when subjected to extensive character-level permutations. To investigate this, we first propose the Scrambled Bench, a suite designed to measure the capacity of LLMs to handle scrambled input, in terms of both recovering scrambled sentences and answering questions given scrambled context. The experimental results indicate that most powerful LLMs demonstrate the capability akin to typoglycemia, a phenomenon where humans can understand the meaning of words even when the letters within those words are scrambled, as long as the first and last letters remain in place. More surprisingly, we found that only GPT-4 nearly flawlessly processes inputs with unnatural errors, even under the extreme condition, a task that poses significant challenges for other LLMs and often even for humans. Specifically, GPT-4 can almost perfectly reconstruct the original sentences from scrambled ones, decreasing the edit distance by 95%, even when all letters within each word are entirely scrambled. It is counter-intuitive that LLMs can exhibit such resilience despite severe disruption to input tokenization caused by scrambled text.