The Ariel Data Challenge Series And Overview of 2025 Edition
Kai Hou Yip
Abstract
This presentation present the Ariel Data Challenge series as a model for connecting mission-driven science with external data science communities. We focus on the 2025 edition, which attracted 1000+ participants from 70+ countries and generated 14000+ submissions. Comparing the 2025 setup with its 2024 predecessor, we analyse how enhanced noise modeling, revised evaluation metrics, and updated observation strategies influenced solution approachees, performance and dominant methodological trends that emerged as a result.
Successful Page Load