2024

A Survey on Hallucination in Large Vision-Language Models

Liu, Hanchao, Xue, Wenyuan, Chen, Yifei et al.

Understand

Recent development of Large Vision-Language Models (LVLMs) has attracted growing attention within the AI landscape for its practical implementation potential.

However, ``hallucination'', or more specifically, the misalignment between factual visual content and corresponding textual generation, poses a significant challenge of utilizing LVLMs.
In this comprehensive survey, we dissect LVLM-related hallucinations in an attempt to establish an overview and facilitate future mitigation.
Our scrutiny starts with a clarification of the concept of hallucinations in LVLMs, presenting a variety of hallucination symptoms and highlighting the unique challenges inherent in LVLM hallucinations.

Built on

Neural baby talk
Jiasen Lu, Jianwei Yang, Dhruv Batra, et al · 2018
Earlier work this paper cites.
Object hallucination in image captioning
Anna Rohrbach, Lisa Anne Hendricks, Kaylee Burns, et al · 2018
Earlier work this paper cites.
The curious case of neural text degeneration
Ari Holtzman, Jan Buys, Li Du, et al · 2020
Earlier work this paper cites.
Learning to summarize with human feedback
Nisan Stiennon, Long Ouyang, Jeffrey Wu, et al · 2020
Earlier work this paper cites.
Neural path hunter: Reducing hallucination in dialogue systems via path grounding
Nouha Dziri, Andrea Madotto, Osmar R Zaiane, et al · 2021
Earlier work this paper cites.
Learning transferable visual models from natural language supervision
Alec Radford, Jong Wook Kim, Chris Hallacy, et al · 2021
Earlier work this paper cites.
Flamingo: a visual language model for few-shot learning
Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, et al · 2022
Earlier work this paper cites.
Fine-grained image captioning with clip reward
Jaemin Cho, Seunghyun Yoon, Ajinkya Kale, et al · 2022
Earlier work this paper cites.
Lora: Low-rank adaptation of large language models
Edward J. Hu, Yelong Shen, Phillip Wallis, et al · 2022
Earlier work this paper cites.
Vision-and-language pretrained models: A survey
Siqu Long, Feiqi Cao, Soyeon Caren Han, et al · 2022
Earlier work this paper cites.
Qwen-vl: A frontier large vision-language model with versatile abilities
Original
Jinze Bai, Shuai Bai, Shusheng Yang, et al · 2023
Earlier work this paper cites.
Position-enhanced visual instruction tuning for multimodal large language models
Original
Chi Chen, Ruoyu Qin, Fuwen Luo, et al · 2023
Earlier work this paper cites.
Minigpt-v2: large language model as a unified interface for vision-language multi-task learning
Original
Jun Chen, Deyao Zhu, Xiaoqian Shen, et al · 2023
Earlier work this paper cites.
Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks
Original
Zhe Chen, Jiannan Wu, Wenhai Wang, et al · 2023
Earlier work this paper cites.
Can we edit multimodal large language models?
Siyuan Cheng, Bozhong Tian, Qingbin Liu, et al · 2023
Earlier work this paper cites.
Dola: Decoding by contrasting layers improves factuality in large language models
Original
Yung-Sung Chuang, Yujia Xie, Hongyin Luo, et al · 2023
Earlier work this paper cites.
Instructblip: Towards general-purpose vision-language models with instruction tuning
Original
Wenliang Dai, Junnan Li, Dongxu Li, et al · 2023
Earlier work this paper cites.
Plausible may not be faithful: Probing object hallucination in vision-language pre-training
Wenliang Dai, Zihan Liu, Ziwei Ji, et al · 2023
Earlier work this paper cites.
Llama-adapter v2: Parameter-efficient visual instruction model
Original
Peng Gao, Jiaming Han, Renrui Zhang, et al · 2023
Earlier work this paper cites.
Imagebind one embedding space to bind them all
Rohit Girdhar, Alaaeldin El-Nouby, Zhuang Liu, et al · 2023
Earlier work this paper cites.

Similar

Detecting and preventing hallucinations in large vision language models
Original
Anisha Gunjal, Jihan Yin, and Erhan Bas · 2023
Cited alongside, same era.
Imagebind-llm: Multi-modality instruction tuning
Original
Jiaming Han, Renrui Zhang, Wenqi Shao, et al · 2023
Cited alongside, same era.
Ciem: Contrastive instruction evaluation method for better instruction tuning
Hongyu Hu, Jiyuan Zhang, Minyi Zhao, et al · 2023
Cited alongside, same era.
A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions
Original
Lei Huang, Weijiang Yu, Weitao Ma, et al · 2023
Cited alongside, same era.
Opera: Alleviating hallucination in multi-modal large language models via over-trust penalty and retrospection-allocation
Original
Qidong Huang, Xiaoyi Dong, Pan Zhang, et al · 2023
Cited alongside, same era.
Vcoder: Versatile vision encoders for multimodal large language models
Original
Jitesh Jain, Jianwei Yang, and Humphrey Shi · 2023
Cited alongside, same era.
Survey of hallucination in natural language generation
Ziwei Ji, Nayeon Lee, Rita Frieske, et al · 2023
Cited alongside, same era.
Hallucination augmented contrastive learning for multimodal large language model
Original
Chaoya Jiang, Haiyang Xu, Mengfan Dong, et al · 2023
Cited alongside, same era.
Faithscore: Evaluating hallucinations in large vision-language models
Original
Liqiang Jing, Ruosen Li, Yunmo Chen, et al · 2023
Cited alongside, same era.
Volcano: Mitigating multimodal hallucination through self-feedback guided revision
Original
Seongyun Lee, Sue Hyun Park, Yongrae Jo, et al · 2023
Cited alongside, same era.
Mitigating object hallucinations in large vision-language models through visual contrastive decoding
Original
Sicong Leng, Hang Zhang, Guanzheng Chen, et al · 2023
Cited alongside, same era.
Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models
Junnan Li, Dongxu Li, Silvio Savarese, et al · 2023
Cited alongside, same era.

Then

Evaluation and mitigation of agnosia in multimodal large language models
Original
Jiaying Lu, Jinmeng Rao, Kezhen Chen, et al · 2023
Later among the works it cites.
Gpt-4 technical report
Original
OpenAI · 2023
Later among the works it cites.
Direct preference optimization: Your language model is secretly a reward model
Original
Rafael Rafailov, Archit Sharma, Eric Mitchell, et al · 2023
Later among the works it cites.
Aligning large multimodal models with factually augmented rlhf
Original
Zhiqing Sun, Sheng Shen, Shengcao Cao, et al · 2023
Later among the works it cites.
Llama: Open and efficient foundation language models
Original
Hugo Touvron, Thibaut Lavril, Gautier Izacard, et al · 2023
Later among the works it cites.
Llama 2: Open foundation and fine-tuned chat models
Original
Hugo Touvron, Louis Martin, Kevin Stone, et al · 2023
Later among the works it cites.
Vigc: Visual instruction generation and correction
Original
Bin Wang, Fan Wu, Xiao Han, et al · 2023
Later among the works it cites.
An llm-free multi-dimensional benchmark for mllms hallucination evaluation
Original
Junyang Wang, Yuhang Wang, Guohai Xu, et al · 2023
Later among the works it cites.
Evaluation and analysis of hallucination in large vision-language models
Original
Junyang Wang, Yiyang Zhou, Guohai Xu, et al · 2023
Later among the works it cites.
A survey on multimodal large language models
Original
Shukang Yin, Chaoyou Fu, Sirui Zhao, et al · 2023
Later among the works it cites.
Woodpecker: Hallucination correction for multimodal large language models
Original
Shukang Yin, Chaoyou Fu, Sirui Zhao, et al · 2023
Later among the works it cites.
Ferret: Refer and ground anything anywhere at any granularity
Original
Haoxuan You, Haotian Zhang, Zhe Gan, et al · 2023
Later among the works it cites.
Rlhf-v: Towards trustworthy mllms via behavior alignment from fine-grained correctional human feedback
Original
Tianyu Yu, Yuan Yao, Haoye Zhang, et al · 2023
Later among the works it cites.
Halle-switch: Rethinking and controlling object existence hallucinations in large vision language models for detailed caption
Original
Bohan Zhai, Shijia Yang, Xiangchen Zhao, et al · 2023
Later among the works it cites.
Recognize anything: A strong image tagging model
Original
Youcai Zhang, Xinyu Huang, Jinyu Ma, et al · 2023
Later among the works it cites.
Siren’s song in the ai ocean: A survey on hallucination in large language models
Original
Yue Zhang, Yafu Li, Leyang Cui, et al · 2023
Later among the works it cites.
Enhancing the spatial awareness capability of multi-modal large language model
Original
Yongqiang Zhao, Zhenyu Li, Zhi Jin, et al · 2023
Later among the works it cites.
Beyond hallucinations: Enhancing lvlms through hallucination-aware direct preference optimization
Original
Zhiyuan Zhao, Bin Wang, Linke Ouyang, et al · 2023
Later among the works it cites.
Analyzing and mitigating object hallucination in large vision-language models
Yiyang Zhou, Chenhang Cui, Jaehong Yoon, et al · 2023
Later among the works it cites.
Minigpt-4: Enhancing vision-language understanding with advanced large language models
Original
Deyao Zhu, Jun Chen, Xiaoqian Shen, et al · 2023
Later among the works it cites.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…