MMSI-Video-Bench: A Benchmark for Video-Based Spatial Intelligence

MMSI-Video-Bench is a holistic benchmark for evaluating spatial intelligence in video-based multimodal models. The dataset, created by author 'rbler', includes video clips and was last updated on February 10, 2026. It is hosted on Hugging Face and has been integrated into the VLMEvalKit framework.

Use Cases

• Benchmarking model performance on video-based spatial intelligence tasks as described in the paper.
• Evaluating multimodal visual question answering capabilities using the provided video clips.
• Training models for multiple-choice question answering based on video content.
• Conducting research on the intersection of video understanding and spatial reasoning.

Strengths

• Designed as a holistic benchmark specifically for video-based spatial intelligence.
• Includes video data in .mp4 format corresponding to each sample.
• Integrated into the VLMEvalKit framework as of December 2025.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file size, and license details are unknown, which may limit suitability assessment.

Provenance

Source

Hugging Face, author rbler

Collection Method

Likely curated for benchmark creation, as indicated by the associated paper and code.

Time Range

null

Freshness

Last updated 2026-02-10 06:01:44

Geography

Region tag suggests 'us', but specific spatial coverage is not detailed.