Autonomous Driving Research Requires
a Community-Driven Data Paradigm

Jinsu Yoo1† Zanming Huang1† Katie Z Luo2 Zheda Mai3 Qiyuan Wu4 Bharath Hariharan4 Mark Campbell4 Wei-Lun Chao1
1Boston University2Stanford University 3The Ohio State University4Cornell University

† Equal contribution

NEURIPS 2026 POSITION PAPER
arXiv · OpenReview ·
Surveyed dataset list: CSV ↓ Markdown ↓
Dataset-paper citation counts are concentrated in a few dominant benchmarks, while many underused datasets cover Indian traffic, Northern European fog and snow, US roadwork, and German heavy rain.
A long tail of underused driving datasets. Citation counts serve as a proxy for dataset usage.
TL;DR

Autonomous driving aims to work anywhere, anytime, for anyone, yet research relies on a narrow set of benchmarks. The community has already collected far more diverse data. Turning these distributed efforts into shared infrastructure can help close the gap between data collection and data utilization.

Explore the 647 dataset papers

Browse the dataset papers used in our literature review. Search by paper title, or select an entry to read its abstract and publication details.

Loading catalog…

Scroll horizontally to see all columns →

Dataset papers matching the title search
Dataset / paper titleYearPaper link