Description

A specialized 100k parameter neural network designed to identify legitimate versus fabricated legal citations. The model weights were 'hand-written' by Claude Fable, achieving 98.9% test accuracy with only ~3,800 non-zero parameters, significantly outperforming traditionally trained models with far more parameters.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.