What type of study is this?

This is a Quantitative Study study.

October 1, 2025Open Access

IV-Bench: A Benchmark for Image-Grounded Video Perception and Reasoning in Multimodal LLMs

View Full Paper

Ask AI

Bookmark

Authors

IV-Bench evaluates image-grounded video reasoning in MLLMs, highlighting significant performance gaps.

Current models achieve only up to 28.9% accuracy on video perception and reasoning tasks.
IV-Bench consists of 967 videos and 2,585 annotated image-text queries covering 13 distinct tasks.
Key performance factors include inference patterns, frame numbers, and resolution in videos.
Findings underscore the need for improved approaches in aligning data formats and refining model training.

Ma et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

October 1, 2025Open Access

View Full Paper

Ask AI

Bookmark

Authors

IV-Bench evaluates image-grounded video reasoning in MLLMs, highlighting significant performance gaps.

Current models achieve only up to 28.9% accuracy on video perception and reasoning tasks.
IV-Bench consists of 967 videos and 2,585 annotated image-text queries covering 13 distinct tasks.
Key performance factors include inference patterns, frame numbers, and resolution in videos.
Findings underscore the need for improved approaches in aligning data formats and refining model training.

Ma et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark