VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Li, Chaoyu, Im, Eun Woo, Fazli, Pooyan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913769043525632
author Li, Chaoyu
Im, Eun Woo
Fazli, Pooyan
author_facet Li, Chaoyu
Im, Eun Woo
Fazli, Pooyan
contents Multimodal large language models (MLLMs) have recently shown significant advancements in video understanding, excelling in content reasoning and instruction-following tasks. However, hallucination, where models generate inaccurate or misleading content, remains underexplored in the video domain. Building on the observation that MLLM visual encoders often fail to distinguish visually different yet semantically similar video pairs, we introduce VidHalluc, the largest benchmark designed to examine hallucinations in MLLMs for video understanding. It consists of 5,002 videos, paired to highlight cases prone to hallucinations. VidHalluc assesses hallucinations across three critical dimensions: (1) action, (2) temporal sequence, and (3) scene transition. Comprehensive testing shows that most MLLMs are vulnerable to hallucinations across these dimensions. Furthermore, we propose DINO-HEAL, a training-free method that reduces hallucinations by incorporating spatial saliency from DINOv2 to reweight visual features during inference. Our results show that DINO-HEAL consistently improves performance on VidHalluc, achieving an average improvement of 3.02% in mitigating hallucinations across all tasks. Both the VidHalluc benchmark and DINO-HEAL code are available at https://people-robots.github.io/vidhalluc.
format Preprint
id arxiv_https___arxiv_org_abs_2412_03735
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding
Li, Chaoyu
Im, Eun Woo
Fazli, Pooyan
Computer Vision and Pattern Recognition
Multimodal large language models (MLLMs) have recently shown significant advancements in video understanding, excelling in content reasoning and instruction-following tasks. However, hallucination, where models generate inaccurate or misleading content, remains underexplored in the video domain. Building on the observation that MLLM visual encoders often fail to distinguish visually different yet semantically similar video pairs, we introduce VidHalluc, the largest benchmark designed to examine hallucinations in MLLMs for video understanding. It consists of 5,002 videos, paired to highlight cases prone to hallucinations. VidHalluc assesses hallucinations across three critical dimensions: (1) action, (2) temporal sequence, and (3) scene transition. Comprehensive testing shows that most MLLMs are vulnerable to hallucinations across these dimensions. Furthermore, we propose DINO-HEAL, a training-free method that reduces hallucinations by incorporating spatial saliency from DINOv2 to reweight visual features during inference. Our results show that DINO-HEAL consistently improves performance on VidHalluc, achieving an average improvement of 3.02% in mitigating hallucinations across all tasks. Both the VidHalluc benchmark and DINO-HEAL code are available at https://people-robots.github.io/vidhalluc.
title VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.03735