Problem
Research questions and friction points this paper is trying to address.
ChildSafeAds
commercial content monitoring
child-facing YouTube videos
data access levels
N\"urnberg NLP at ChildSafeAds 2026: Structurally Dissimilar Voter Ensembles under Four Levels of Data Access
🤖 AI Summary
The provided TLDR content is incomplete, consisting solely of fragmented phrases such as "this project adopts" and "this project uses," while lacking essential information regarding the specific research problem, technical methodology, and core contributions of the paper. Consequently, direct optimization is not feasible. To proceed, please supply a complete project abstract or key details from the original manuscript. Upon receipt of the necessary information, the summary will be refined in strict accordance with the standard academic structure encompassing the problem, method, and contributions or results. The revised text will be condensed to 150–200 words, ensuring that the primary innovations are clearly articulated and emphasized.
📝 Abstract
We describe the N\"urnberg NLP system for ChildSafeAds 2026. The shared task asks what a monitoring system for commercial content in child-facing YouTube videos can achieve at a given level of data access. We answer with per-subtask ensembles of nine voters, organised into three branches that differ in backbone, adaptation method and class scope. Selection rests on channel-disjoint cross-validation, with the development set as a transfer check. The system wins two of the three subtasks. Its product-category score (ST2, 0.8243) and its compliance-flag score (ST3, 0.6530) are the best of the 22 final entries, and it places third on the task mean (0.7079). We further compare four access levels and report the cost at test-set scale.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Structurally Dissimilar Ensembles
Channel-Disjoint Cross-Validation
Child-Facing Content Monitoring
Data Access Levels
Voter Ensembles
🔎 Similar Papers
No similar papers found.
