Datasets:

zhehuderek
/

VIVA_Plus_Benchmark

Modalities:

Formats:

Size:

ArXiv:

Libraries:

License:

Dataset card Data Studio Files Files and versions

Dataset Viewer

The dataset viewer is not available for this dataset.

Unexpected token '<', "<html> <h"... is not valid JSON

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

VIVA+: Human-Centered Situational Decision-Making

Zhe Hu¹, Yixiao Ren¹, Guanzhong Liu¹, Jing Li^1,2, Yu Yin³,

¹The Hong Kong Polytechnic University

²Research Centre for Data Science & Artificial Intelligence

³Case Western Reserve University

[Webpage] [arXiv] [GitHub]

Overview

We introduce VIVA+, a cognitively grounded benchmark for evaluating the reasoning and decision-making of MLLMs in human-centered situations. VIVA+ consists of 1,317 real-world situations paired with 6,373 multiple-choice questions, targeting three core abilities for decision-making: (1) Foundational Situation Comprehension, (2) Context-Driven Action Justification, and (3) Reflective Reasoning.

Downloads last month: 66

Total file size:

Paper for zhehuderek/VIVA_Plus_Benchmark

VIVA+: Human-Centered Situational Decision-Making

Paper • 2509.23698 • Published Sep 28, 2025 • 1