Legal Judgment Prediction
Overview
The field studies how well computational systems can predict decisions such as verdicts, charges, sentences, and damages from legal documents and case details. Approaches range from feature-based and neural text classifiers to transformer language models, retrieval-augmented systems, and methods that structure reasoning around legal rules, causal factors, or interactions among claims. Benchmarks such as CAIL2018, COLIEE, and ECtHR case datasets test outcomes and related tasks, but strong scores do not always show that a model relied on the law or evidence in a meaningful way. Key challenges include temporal change, biased or outcome-leaking data, uncertain labels, weak calibration, and explanations that cite legal authorities without actually depending on them.
Latest papers
29 papers3 papers in the last four weeks, down 25% on the four weeks before. 0.0% of all new papers.