E-Bench: Towards Evaluating the Ease-of-Use of Large Language Models

TL;DR

This paper introduces E-Bench, a benchmark for evaluating the robustness and ease-of-use of large language models against prompt perturbations like paraphrasing, simplification, colloquialism, and typos, revealing that larger models are more robust but still not user-friendly enough.

Contribution

The paper presents E-Bench, a systematic benchmark for assessing LLMs' stability to prompt perturbations, addressing a gap in evaluating model robustness in real-world scenarios.

Findings

Larger models show improved robustness to prompt perturbations

Prompt perturbations significantly degrade LLM performance

There is still a considerable gap in making LLMs user-friendly

Abstract

Most large language models (LLMs) are sensitive to prompts, and another synonymous expression or a typo may lead to unexpected results for the model. Composing an optimal prompt for a specific demand lacks theoretical support and relies entirely on human experimentation, which poses a considerable obstacle to popularizing generative artificial intelligence. However, there is no systematic analysis of the stability of LLMs in resisting prompt perturbations in real-world scenarios. In this work, we propose to evaluate the ease-of-use of LLMs and construct E-Bench, simulating the actual situation of human use from synonymous perturbation (including paraphrasing, simplification, and colloquialism) and typographical perturbation (such as typing). On this basis, we also discuss the combination of these two types of perturbation and analyze the main reasons for performance degradation. Experimental results indicate that with the increase of model size, although the ease-of-use are significantly improved, there is still a long way to go to build a sufficiently user-friendly model.