Hallu-PI: Evaluating Hallucination in Multi-modal Large Language Models within Perturbed Inputs

TL;DR

Hallu-PI is a new benchmark that evaluates hallucination in multi-modal large language models when inputs are perturbed, revealing significant hallucination issues not seen in unperturbed scenarios and highlighting the need for more robust models.

Contribution

This paper introduces Hallu-PI, the first benchmark for assessing hallucination in MLLMs under input perturbations, along with baseline methods for such scenarios.

Findings

MLLMs exhibit increased hallucination on perturbed inputs.

Models show bias towards certain hallucination types.

Perturbed-Reminder and Perturbed-ICL improve robustness.

Abstract

Multi-modal Large Language Models (MLLMs) have demonstrated remarkable performance on various visual-language understanding and generation tasks. However, MLLMs occasionally generate content inconsistent with the given images, which is known as "hallucination". Prior works primarily center on evaluating hallucination using standard, unperturbed benchmarks, which overlook the prevalent occurrence of perturbed inputs in real-world scenarios-such as image cropping or blurring-that are critical for a comprehensive assessment of MLLMs' hallucination. In this paper, to bridge this gap, we propose Hallu-PI, the first benchmark designed to evaluate Hallucination in MLLMs within Perturbed Inputs. Specifically, Hallu-PI consists of seven perturbed scenarios, containing 1,260 perturbed images from 11 object types. Each image is accompanied by detailed annotations, which include fine-grained hallucination types, such as existence, attribute, and relation. We equip these annotations with a rich set of questions, making Hallu-PI suitable for both discriminative and generative tasks. Extensive experiments on 12 mainstream MLLMs, such as GPT-4V and Gemini-Pro Vision, demonstrate that these models exhibit significant hallucinations on Hallu-PI, which is not observed in unperturbed scenarios. Furthermore, our research reveals a severe bias in MLLMs' ability to handle different types of hallucinations. We also design two baselines specifically for perturbed scenarios, namely Perturbed-Reminder and Perturbed-ICL. We hope that our study will bring researchers' attention to the limitations of MLLMs when dealing with perturbed inputs, and spur further investigations to address this issue. Our code and datasets are publicly available at https://github.com/NJUNLP/Hallu-PI.