KITScenes LongTail: Multi-View Driving Data for Rare Scenario Generalization

KITScenes LongTail is a dataset for end-to-end driving research focusing on long-tail events. It provides multi-view video data, vehicle trajectories, high-level instructions, and detailed reasoning traces. The dataset was created by KIT-MRT and was last updated on Hugging Face in April 2026.

Use Cases

• Benchmarking few-shot generalization of multimodal models based on the provided reasoning traces.
• Training models for trajectory prediction based on multi-view video data.
• Studying in-context learning for autonomous driving using high-level instructions.
• Developing systems to handle rare driving scenarios based on the long-tail event focus.

Strengths

• Focuses on long-tail driving events, a key challenge for real-world generalization.
• Includes multiple data modalities: video, trajectories, instructions, and reasoning traces.
• Designed to facilitate in-context learning and few-shot generalization.

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 and file size are unknown, which may limit suitability assessment.

Provenance

Source

KIT-MRT

Collection Method

Introduced as a new dataset for end-to-end driving research.

Time Range

null

Freshness

Last updated 2026-04-01 10:48:49; freshness should be verified.

Geography

null