01 Overview
Understanding complex driving scenarios requires more than object detection. It demands reasoning about the causal chains that connect agents, actions, and environmental context over time. The AV Causal Scenario Retrieval Challenge invites the research community to develop models that can retrieve driving scenarios from dense, structured annotations capturing spatial, temporal, and causal relationships.
The challenge is built on the CASCADE dataset, which provides dense human annotations for NVIDIA's public PAI-AV Dataset. CASCADE captures hierarchical, temporally dense relationships across real-world dashcam video including multi-entity-class schemas, explicit causal links, spatial containment hierarchies, and normative compliance flags.
02 Challenge Track
Details on challenge tracks, submission format, evaluation protocol, and scoring will be announced soon.
Causal Scenario Retrieval
Task: Given a natural-language query describing a complex driving situation involving spatial, temporal, and/or causal relationships, retrieve the matching scenario(s) from the annotated corpus.
Full task specification and baseline details will be published soon.
03 Timeline
- 2026-06-03 Challenge announcement
- 2026-07-31 Competition goes live Registration, public data, starter tools, and submission instructions open.
- 2026-11-20 Public leaderboard closes Final submissions accepted for the leaderboard.
- 2026-11-29 Final results released Final results and winners are released to participants.
- NeurIPS 2026 Public winner announcement at NeurIPS 2026 Top entries presented at a workshop session.
Dates are tentative and subject to update.
04 Organizers
This challenge is organized by NVIDIA's Spatial Intelligence Lab.
HostNVIDIA Spatial Intelligence Lab
NVIDIA Research lab working on spatial intelligence for autonomous systems, 3D understanding, and scene reasoning.