Multimodal Safety Benchmark for Autonomous Driving

SafeBench and its subset MiniBench provide multimodal data for safety benchmarking. The datasets contain text, audio, and image modalities, with MiniBench being a random sample of SafeBench. Author Zonghao2025 published the collection in September 2024.

Use Cases

• Train multimodal models on the text, audio, and image data for autonomous driving safety evaluations.
• Benchmark model robustness using the safety-focused labels implied by the 'Safety Benchmarking' tag.
• Analyze the relationship between different data modalities (text, audio, image) within the context of autonomous systems.

Strengths

• Dataset was last updated in September 2024, indicating recent maintenance.
• Covers three distinct data modalities: text, audio, and images.
• Explicitly tagged for 'Safety Benchmarking' and 'Autonomous Driving', providing clear application context.

Limitations

• Unknown dataset size, row count, and column structure prevents assessment of scale and granularity.
• Sample data and file formats are unavailable, making it difficult to evaluate data quality and structure before download.
• The description is incomplete, requiring users to visit an external page for full details.

Provenance

Source

huggingface

Collection Method

Repository provides two datasets: MiniBench and SafeBench, with the former randomly sampled from the latter.

Freshness

Last updated 2024-09-28.