Assessing Modality Bias in Video Question Answering Benchmarks with Multimodal Large Language Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Park, Jean, Jang, Kuk Jin, Alasaly, Basam, Mopidevi, Sriharsha, Zolensky, Andrew, Eaton, Eric, Lee, Insup, Johnson, Kevin
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912162422718464
author Park, Jean
Jang, Kuk Jin
Alasaly, Basam
Mopidevi, Sriharsha
Zolensky, Andrew
Eaton, Eric
Lee, Insup
Johnson, Kevin
author_facet Park, Jean
Jang, Kuk Jin
Alasaly, Basam
Mopidevi, Sriharsha
Zolensky, Andrew
Eaton, Eric
Lee, Insup
Johnson, Kevin
contents Multimodal large language models (MLLMs) can simultaneously process visual, textual, and auditory data, capturing insights that complement human analysis. However, existing video question-answering (VidQA) benchmarks and datasets often exhibit a bias toward a single modality, despite the goal of requiring advanced reasoning skills that integrate diverse modalities to answer the queries. In this work, we introduce the modality importance score (MIS) to identify such bias. It is designed to assess which modality embeds the necessary information to answer the question. Additionally, we propose an innovative method using state-of-the-art MLLMs to estimate the modality importance, which can serve as a proxy for human judgments of modality perception. With this MIS, we demonstrate the presence of unimodal bias and the scarcity of genuinely multimodal questions in existing datasets. We further validate the modality importance score with multiple ablation studies to evaluate the performance of MLLMs on permuted feature sets. Our results indicate that current models do not effectively integrate information due to modality imbalance in existing datasets. Our proposed MLLM-derived MIS can guide the curation of modality-balanced datasets that advance multimodal learning and enhance MLLMs' capabilities to understand and utilize synergistic relations across modalities.
format Preprint
id arxiv_https___arxiv_org_abs_2408_12763
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Assessing Modality Bias in Video Question Answering Benchmarks with Multimodal Large Language Models
Park, Jean
Jang, Kuk Jin
Alasaly, Basam
Mopidevi, Sriharsha
Zolensky, Andrew
Eaton, Eric
Lee, Insup
Johnson, Kevin
Machine Learning
Artificial Intelligence
Computation and Language
Multimodal large language models (MLLMs) can simultaneously process visual, textual, and auditory data, capturing insights that complement human analysis. However, existing video question-answering (VidQA) benchmarks and datasets often exhibit a bias toward a single modality, despite the goal of requiring advanced reasoning skills that integrate diverse modalities to answer the queries. In this work, we introduce the modality importance score (MIS) to identify such bias. It is designed to assess which modality embeds the necessary information to answer the question. Additionally, we propose an innovative method using state-of-the-art MLLMs to estimate the modality importance, which can serve as a proxy for human judgments of modality perception. With this MIS, we demonstrate the presence of unimodal bias and the scarcity of genuinely multimodal questions in existing datasets. We further validate the modality importance score with multiple ablation studies to evaluate the performance of MLLMs on permuted feature sets. Our results indicate that current models do not effectively integrate information due to modality imbalance in existing datasets. Our proposed MLLM-derived MIS can guide the curation of modality-balanced datasets that advance multimodal learning and enhance MLLMs' capabilities to understand and utilize synergistic relations across modalities.
title Assessing Modality Bias in Video Question Answering Benchmarks with Multimodal Large Language Models
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2408.12763