VORD: Visual Ordinal Calibration for Mitigating Object Hallucinations in Large Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Neo, Dexter, Chen, Tsuhan
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915073261305856
author Neo, Dexter
Chen, Tsuhan
author_facet Neo, Dexter
Chen, Tsuhan
contents Large Vision-Language Models (LVLMs) have made remarkable developments along with the recent surge of large language models. Despite their advancements, LVLMs have a tendency to generate plausible yet inaccurate or inconsistent information based on the provided source content. This phenomenon, also known as ``hallucinations" can have serious downstream implications during the deployment of LVLMs. To address this, we present VORD a simple and effective method that alleviates hallucinations by calibrating token predictions based on ordinal relationships between modified image pairs. VORD is presented in two forms: 1.) a minimalist training-free variant which eliminates implausible tokens from modified image pairs, and 2.) a trainable objective function that penalizes unlikely tokens. Our experiments demonstrate that VORD delivers better calibration and effectively mitigates object hallucinations on a wide-range of LVLM benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2412_15739
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle VORD: Visual Ordinal Calibration for Mitigating Object Hallucinations in Large Vision-Language Models
Neo, Dexter
Chen, Tsuhan
Computer Vision and Pattern Recognition
Large Vision-Language Models (LVLMs) have made remarkable developments along with the recent surge of large language models. Despite their advancements, LVLMs have a tendency to generate plausible yet inaccurate or inconsistent information based on the provided source content. This phenomenon, also known as ``hallucinations" can have serious downstream implications during the deployment of LVLMs. To address this, we present VORD a simple and effective method that alleviates hallucinations by calibrating token predictions based on ordinal relationships between modified image pairs. VORD is presented in two forms: 1.) a minimalist training-free variant which eliminates implausible tokens from modified image pairs, and 2.) a trainable objective function that penalizes unlikely tokens. Our experiments demonstrate that VORD delivers better calibration and effectively mitigates object hallucinations on a wide-range of LVLM benchmarks.
title VORD: Visual Ordinal Calibration for Mitigating Object Hallucinations in Large Vision-Language Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.15739