PaSBench-Video: A Benchmark for Proactive Safety and Risk Anticipation

PaSBench-Video is a curated benchmark containing 481 video files for proactive safety and risk anticipation. Each sample pairs a video with frame/time annotations and reviewed risk-response prompts. The dataset was created by beingbetter11643 and was last updated on Hugging Face in May 2026.

Use Cases

• Training models for proactive safety anticipation based on video and time annotations.
• Benchmarking risk-response generation models using the reviewed prompts.
• Developing multimodal AI systems that integrate visual and temporal cues for safety.

Strengths

• Contains 481 curated video files.
• Includes frame/time annotations and reviewed prompts for each sample.
• Last updated on Hugging Face in May 2026.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect source bias inherent to the original video datasets.

Provenance

Source

beingbetter11643 on Hugging Face

Collection Method

Curated from multiple original source datasets.

Freshness

2026-05-28