Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems

Tolga Dimlioglu 1, Nadine Chang 2, Maying Shen 2, Rafid Mahmood 2,3, Jose M. Alvarez 2 · 2026

Read Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems by Tolga Dimlioglu 1, Nadine Chang 2, Maying Shen 2, Rafid Mahmood 2,3, Jose M. Alvarez 2 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-08366.txt.

Description

Tolga Dimlioglu and four colleagues tackle training data choice for end to end driving systems, where one data mix must serve many scoring rules at once. Their MOSAIC method splits data into domains, fits scaling laws from each domain to each score, and grows the mix by adding data where scores gain most. On a driving planner score it beat varied rivals while using up to 80 percent less data.

Catalog record

CREATORTolga Dimlioglu 1, Nadine Chang 2, Maying Shen 2, Rafid Mahmood 2,3, Jose M. Alvarez 2
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73759
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.08366

Browse the full library · About the Tower of Babel