ObfusQAte: A Proposed Framework to Evaluate LLM Robustness on Obfuscated Factual Question Answering

TL;DR

This paper introduces ObfusQA, a novel framework for systematically evaluating the robustness of large language models against obfuscated factual questions across multiple complexity levels.

Contribution

It presents the first comprehensive benchmark, ObfusQA, with multi-tiered obfuscation techniques to assess LLM performance and robustness in challenging question scenarios.

Findings

LLMs often fail or hallucinate with obfuscated questions

ObfusQA reveals vulnerabilities in LLM robustness

Framework is publicly available for future research

Abstract

The rapid proliferation of Large Language Models (LLMs) has significantly contributed to the development of equitable AI systems capable of factual question-answering (QA). However, no known study tests the LLMs' robustness when presented with obfuscated versions of questions. To systematically evaluate these limitations, we propose a novel technique, ObfusQAte, and leveraging the same, introduce ObfusQA, a comprehensive, first-of-its-kind framework with multi-tiered obfuscation levels designed to examine LLM capabilities across three distinct dimensions: (i) Named-Entity Indirection, (ii) Distractor Indirection, and (iii) Contextual Overload. By capturing these fine-grained distinctions in language, ObfusQA provides a comprehensive benchmark for evaluating LLM robustness and adaptability. Our study observes that LLMs exhibit a tendency to fail or generate hallucinated responses when confronted with these increasingly nuanced variations. To foster research in this direction, we make ObfusQAte publicly available.