MuG: A Multimodal Classification Benchmark on Game Data with Tabular, Textual, and Visual Fields

TL;DR

MuG is a comprehensive multimodal game data benchmark with eight datasets across different genres, enabling evaluation and development of models that integrate tabular, textual, and visual information.

Contribution

This paper introduces MuG, a new benchmark with diverse datasets for multimodal classification in gaming, along with analysis and baseline results.

Findings

Multimodal classifiers outperform unimodal ones on MuG.

The benchmark reveals the complexity and dependency of multimodal data.

MuG provides a challenging platform for multimodal model development.

Abstract

Previous research has demonstrated the advantages of integrating data from multiple sources over traditional unimodal data, leading to the emergence of numerous novel multimodal applications. We propose a multimodal classification benchmark MuG with eight datasets that allows researchers to evaluate and improve their models. These datasets are collected from four various genres of games that cover tabular, textual, and visual modalities. We conduct multi-aspect data analysis to provide insights into the benchmark, including label balance ratios, percentages of missing features, distributions of data within each modality, and the correlations between labels and input modalities. We further present experimental results obtained by several state-of-the-art unimodal classifiers and multimodal classifiers, which demonstrate the challenging and multimodal-dependent properties of the benchmark. MuG is released at https://github.com/lujiaying/MUG-Bench with the data, tutorials, and implemented baselines.