Probing Visual Language Priors in VLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Luo, Tiange, Cao, Ang, Lee, Gunhee, Johnson, Justin, Lee, Honglak
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909575953776640
author Luo, Tiange
Cao, Ang
Lee, Gunhee
Johnson, Justin
Lee, Honglak
author_facet Luo, Tiange
Cao, Ang
Lee, Gunhee
Johnson, Justin
Lee, Honglak
contents Despite recent advances in Vision-Language Models (VLMs), they may over-rely on visual language priors existing in their training data rather than true visual reasoning. To investigate this, we introduce ViLP, a benchmark featuring deliberately out-of-distribution images synthesized via image generation models and out-of-distribution Q&A pairs. Each question in ViLP is coupled with three potential answers and three corresponding images: one that can be resolved by text priors alone and two that demand visual reasoning. Although, humans achieve near-perfect accuracy, modern VLMs falter; for instance, GPT-4 achieves only 66.17% on ViLP. To alleviate this, we propose a self-improving framework in which models generate new VQA data, then apply pixel-level and semantic corruptions to form "good-bad" image pairs for self-training. Our training objectives compel VLMs to focus more on the actual visual inputs, and we demonstrate their effectiveness in boosting the performance of open-source VLMs, including LLaVA-v1.5 and Cambrian.
format Preprint
id arxiv_https___arxiv_org_abs_2501_00569
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Probing Visual Language Priors in VLMs
Luo, Tiange
Cao, Ang
Lee, Gunhee
Johnson, Justin
Lee, Honglak
Computer Vision and Pattern Recognition
Machine Learning
Despite recent advances in Vision-Language Models (VLMs), they may over-rely on visual language priors existing in their training data rather than true visual reasoning. To investigate this, we introduce ViLP, a benchmark featuring deliberately out-of-distribution images synthesized via image generation models and out-of-distribution Q&A pairs. Each question in ViLP is coupled with three potential answers and three corresponding images: one that can be resolved by text priors alone and two that demand visual reasoning. Although, humans achieve near-perfect accuracy, modern VLMs falter; for instance, GPT-4 achieves only 66.17% on ViLP. To alleviate this, we propose a self-improving framework in which models generate new VQA data, then apply pixel-level and semantic corruptions to form "good-bad" image pairs for self-training. Our training objectives compel VLMs to focus more on the actual visual inputs, and we demonstrate their effectiveness in boosting the performance of open-source VLMs, including LLaVA-v1.5 and Cambrian.
title Probing Visual Language Priors in VLMs
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2501.00569