Visual Description Grounding Reduces Hallucinations and Boosts Reasoning in LVLMs

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ghosh, Sreyan, Evuru, Chandra Kiran Reddy, Kumar, Sonal, Tyagi, Utkarsh, Nieto, Oriol, Jin, Zeyu, Manocha, Dinesh
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913720886624256
author Ghosh, Sreyan
Evuru, Chandra Kiran Reddy
Kumar, Sonal
Tyagi, Utkarsh
Nieto, Oriol
Jin, Zeyu
Manocha, Dinesh
author_facet Ghosh, Sreyan
Evuru, Chandra Kiran Reddy
Kumar, Sonal
Tyagi, Utkarsh
Nieto, Oriol
Jin, Zeyu
Manocha, Dinesh
contents Large Vision-Language Models (LVLMs) often produce responses that misalign with factual information, a phenomenon known as hallucinations. While hallucinations are well-studied, the exact causes behind them remain underexplored. In this paper, we first investigate the root causes of hallucinations in LVLMs. Our findings reveal that existing mitigation techniques primarily reduce hallucinations for visual recognition prompts-those that require simple descriptions of visual elements-but fail for cognitive prompts that demand deliberate reasoning. We identify the core issue as a lack of true visual perception in LVLMs: although they can accurately recognize visual elements, they struggle to fully interpret these elements in the context of the input prompt and effectively link this recognition to their internal knowledge, which is critical for reasoning. To address this gap, we introduce Visual Description Grounded Decoding (VDGD), a simple, robust, and training-free method designed to enhance visual perception and improve reasoning capabilities in LVLMs. VDGD works by first generating a detailed description of the image and appending it as a prefix to the instruction. During response generation, tokens are sampled based on their KL divergence to the description, favoring candidates with lower divergence. Experimental results on multiple visual reasoning benchmarks and LVLMs demonstrate that VDGD consistently outperforms existing baselines 2% - 33%. Finally, we introduce VaLLu, a benchmark designed for comprehensive evaluation of the cognitive capabilities of LVLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2405_15683
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Visual Description Grounding Reduces Hallucinations and Boosts Reasoning in LVLMs
Ghosh, Sreyan
Evuru, Chandra Kiran Reddy
Kumar, Sonal
Tyagi, Utkarsh
Nieto, Oriol
Jin, Zeyu
Manocha, Dinesh
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Large Vision-Language Models (LVLMs) often produce responses that misalign with factual information, a phenomenon known as hallucinations. While hallucinations are well-studied, the exact causes behind them remain underexplored. In this paper, we first investigate the root causes of hallucinations in LVLMs. Our findings reveal that existing mitigation techniques primarily reduce hallucinations for visual recognition prompts-those that require simple descriptions of visual elements-but fail for cognitive prompts that demand deliberate reasoning. We identify the core issue as a lack of true visual perception in LVLMs: although they can accurately recognize visual elements, they struggle to fully interpret these elements in the context of the input prompt and effectively link this recognition to their internal knowledge, which is critical for reasoning. To address this gap, we introduce Visual Description Grounded Decoding (VDGD), a simple, robust, and training-free method designed to enhance visual perception and improve reasoning capabilities in LVLMs. VDGD works by first generating a detailed description of the image and appending it as a prefix to the instruction. During response generation, tokens are sampled based on their KL divergence to the description, favoring candidates with lower divergence. Experimental results on multiple visual reasoning benchmarks and LVLMs demonstrate that VDGD consistently outperforms existing baselines 2% - 33%. Finally, we introduce VaLLu, a benchmark designed for comprehensive evaluation of the cognitive capabilities of LVLMs.
title Visual Description Grounding Reduces Hallucinations and Boosts Reasoning in LVLMs
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2405.15683