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

arXiv: https://arxiv.org/abs/2408.01355 · HTML: https://mlsanity.com/paper/2408.01355/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Hallucination mitigation in LVLMs, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 20 of its 36 references that are in the map.
- First posted on arXiv 2 Aug 2024.

## Placed in

- Multimodal LLMs / vision-language models / [Hallucination mitigation in LVLMs](https://mlsanity.com/mllm/hallucination-mitigation-in-lvlms/)

## Builds on (most cited first, 20 of 20)

- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Benchmarking Neural Network Robustness to Common Corruptions and Perturbations](https://mlsanity.com/paper/1903.12261/) (2019)
- [RandAugment: Practical automated data augmentation with a reduced search space](https://mlsanity.com/paper/1909.13719/) (2019)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [Survey of Hallucination in Natural Language Generation](https://mlsanity.com/paper/2202.03629/) (2022)
- [PaLM-E: An Embodied Multimodal Language Model](https://mlsanity.com/paper/2303.03378/) (2023)
- [MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models](https://mlsanity.com/paper/2306.13394/) (2023)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [GLM: General Language Model Pretraining with Autoregressive Blank Infilling](https://mlsanity.com/paper/2103.10360/) (2021)
- [SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension](https://mlsanity.com/paper/2307.16125/) (2023)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [A Survey on In-context Learning](https://mlsanity.com/paper/2301.00234/) (2022)
- [mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration](https://mlsanity.com/paper/2311.04257/) (2023)
- [OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocation](https://mlsanity.com/paper/2311.17911/) (2023)
- [Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming](https://mlsanity.com/paper/1907.07484/) (2019)
- [Analyzing and Mitigating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2310.00754/) (2023)
- [InternLM-XComposer: A Vision-Language Large Model for Advanced Text-image Comprehension and Composition](https://mlsanity.com/paper/2309.15112/) (2023)
- [A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily](https://mlsanity.com/paper/2311.08268/) (2023)
- [MultiModal-GPT: A Vision and Language Model for Dialogue with Humans](https://mlsanity.com/paper/2305.04790/) (2023)
- [Red Teaming GPT-4V: Are GPT-4V Safe Against Uni/Multi-Modal Jailbreak Attacks?](https://mlsanity.com/paper/2404.03411/) (2024)

## Built on by (most cited first, 14 of 14)

- [A Survey of State of the Art Large Vision Language Models: Alignment, Benchmark, Evaluations and Challenges](https://mlsanity.com/paper/2501.02189/) (2025)
- [A Survey on Multimodal Benchmarks: In the Era of Large AI Models](https://mlsanity.com/paper/2409.18142/) (2024)
- [VidHal: Benchmarking Temporal Hallucinations in Vision LLMs](https://mlsanity.com/paper/2411.16771/) (2024)
- [Measuring Epistemic Humility in Multimodal Large Language Models](https://mlsanity.com/paper/2509.09658/) (2025)
- [FINER: MLLMs Hallucinate under Fine-grained Negative Queries](https://mlsanity.com/paper/2603.17662/) (2026)
- [Can Humans Dream of Electric Sheep? Human-Written Samples for Fine-Grained Vision-and-Language Hallucination Benchmarking](https://mlsanity.com/paper/2608.01021/) (2026)
- [HIVTP: A Training-Free Method to Improve VLMs Efficiency via Hierarchical Visual Token Pruning Using Middle-Layer-Based Importance Score](https://mlsanity.com/paper/2509.23663/) (2025)
- [SCAN: Structured Capability Assessment and Navigation for LLMs](https://mlsanity.com/paper/2505.06698/) (2025)
- [KnowHal: A Knowledge-Driven Benchmark for Comprehensive Multimodal Hallucination Evaluation](https://mlsanity.com/paper/2608.03782/) (2026)
- [SHALE: A Scalable Benchmark for Fine-grained Hallucination Evaluation in LVLMs](https://mlsanity.com/paper/2508.09584/) (2025)
- [FaithSCAN: Model-Driven Single-Pass Hallucination Detection for Faithful Visual Question Answering](https://mlsanity.com/paper/2601.00269/) (2026)
- [Unified Hallucination Fuzzing for Multimodal Large Language Models](https://mlsanity.com/paper/2608.07525/) (2026)
- [Contextualized Evaluation of Vision Language Models through Dynamic, Multi-turn Interactions](https://mlsanity.com/paper/2607.14499/) (2026)
- [Overview of SHROOM-Visions 2026: A Shared Task on Hallucination Detection in Large Vision-Language Models](https://mlsanity.com/paper/2608.25662/) (2026)

## Abstract (arXiv)

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.