Salvato in:
Dettagli Bibliografici
Autori principali: Yu, Le, Wang, Kaishen, Xiong, Jianlong, Cao, Yue, Zhang, Lei, He, Zhang Yi Tao
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2506.17587
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910007027564544
author Yu, Le
Wang, Kaishen
Xiong, Jianlong
Cao, Yue
Zhang, Lei
He, Zhang Yi Tao
author_facet Yu, Le
Wang, Kaishen
Xiong, Jianlong
Cao, Yue
Zhang, Lei
He, Zhang Yi Tao
contents Though Large Vision-Language Models (LVLMs) have achieved remarkable performance across various tasks, they are still prone to hallucinations-generating outputs that are textually plausible but visually ungrounded. While prior approaches generally address this issue through data-centric fine-tuning or innovative decoding strategies, these methods often require substantial resources or task-specific configurations. In this work, we introduce an architecture-level solution, HalluRNN, which enhances model stability through recurrent cross-layer reasoning. Specifically, we propose a novel Dual-Gated Depth Propagation Unit (DG-DPU) module, which is shared across layers and recurrently refines hidden states. This allows for the adaptive propagation of information throughout the model, enforces consistency across layers, and mitigates hallucinations caused by representational drift. By fine-tuning only the DG-DPU module, HalluRNN achieves strong and robust performance across multiple benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_17587
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HalluRNN: Mitigating Hallucinations via Recurrent Cross-Layer Reasoning in Large Vision-Language Models
Yu, Le
Wang, Kaishen
Xiong, Jianlong
Cao, Yue
Zhang, Lei
He, Zhang Yi Tao
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Though Large Vision-Language Models (LVLMs) have achieved remarkable performance across various tasks, they are still prone to hallucinations-generating outputs that are textually plausible but visually ungrounded. While prior approaches generally address this issue through data-centric fine-tuning or innovative decoding strategies, these methods often require substantial resources or task-specific configurations. In this work, we introduce an architecture-level solution, HalluRNN, which enhances model stability through recurrent cross-layer reasoning. Specifically, we propose a novel Dual-Gated Depth Propagation Unit (DG-DPU) module, which is shared across layers and recurrently refines hidden states. This allows for the adaptive propagation of information throughout the model, enforces consistency across layers, and mitigates hallucinations caused by representational drift. By fine-tuning only the DG-DPU module, HalluRNN achieves strong and robust performance across multiple benchmarks.
title HalluRNN: Mitigating Hallucinations via Recurrent Cross-Layer Reasoning in Large Vision-Language Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2506.17587