AnnoCaseLaw: Legal Annotation Benchmark of Civil Negligence
AnnoCaseLaw is a new dataset of 471 U.S. Appeals Court civil negligence cases with over 12,000 expert annotations across five categories like Facts and Outcome. The study tested frontier AI models on predicting judicial decisions and identifying legal concepts. Results show that applying the law to facts is hard for models, emphasizing the need for high-quality data in legal AI.