Form Over Content In Gradient-Based Data Attribution Methods

A technical paper on gradient-based data attribution methods, with no direct AI risk or safety implications.

Auto-scored: This is a methodological machine learning paper with no risk or mitigation relevance to the ARQ rubric.