GameplayQA Benchmark For Multi-Agent Video Understanding

GameplayQA is the first benchmark for POV-Synced Multi-Video Understanding and Multi-Agent Video tasks. The dataset contains video, text, and tabular modalities for gameplay understanding and visual question answering. It was created by researchers at the University of Southern California and published in 2026.

Use Cases

• Train models for Video Question Answering using synchronized multi-video inputs.
• Benchmark Multi-Agent Video Understanding systems on POV-synced gameplay data.
• Develop Multi-Video Understanding models for tasks like Video Text To Text generation.
• Analyze Gameplay Understanding through the provided visual and textual modalities.

Strengths

• First benchmark for POV-Synced Multi-Video Understanding, a novel task.
• Includes multiple modalities: video, text, and tabular data.
• Dataset was updated in March 2026, indicating recent maintenance.

Limitations

• Specific scale details like row count, column count, and video duration are unknown.
• Potential bias towards 3D virtual agent environments and US region data.
• Label quality and annotation methodology are not described in the provided input.

Provenance

Source

Hugging Face dataset uploaded by author 'wangyz1999'.

Collection Method

null

Time Range

null

Freshness

Last updated on 2026-03-24.

Geography

Region tag indicates 'us'.