Gaps or Hallucinations? Gazing into Machine-Generated Legal Analysis for Fine-grained Text Evaluations

TL;DR

This paper investigates the evaluation of machine-generated legal analysis by introducing the concept of gaps, a nuanced alternative to hallucinations, and develops a detector to identify these gaps in LLM outputs, revealing high hallucination rates.

Contribution

It introduces the notion of gaps in legal analysis, creates a taxonomy and a detector for gap categories, and provides an annotated dataset for automatic evaluation of LLM-generated legal texts.

Findings

The best detector achieves 67% F1 score and 80% precision.

Approximately 80% of LLM-generated legal analyses contain hallucinations.

Gaps are a nuanced form of errors that do not always imply invalidity.

Abstract

Large Language Models (LLMs) show promise as a writing aid for professionals performing legal analyses. However, LLMs can often hallucinate in this setting, in ways difficult to recognize by non-professionals and existing text evaluation metrics. In this work, we pose the question: when can machine-generated legal analysis be evaluated as acceptable? We introduce the neutral notion of gaps, as opposed to hallucinations in a strict erroneous sense, to refer to the difference between human-written and machine-generated legal analysis. Gaps do not always equate to invalid generation. Working with legal experts, we consider the CLERC generation task proposed in Hou et al. (2024b), leading to a taxonomy, a fine-grained detector for predicting gap categories, and an annotated dataset for automatic evaluation. Our best detector achieves 67% F1 score and 80% precision on the test set. Employing this detector as an automated metric on legal analysis generated by SOTA LLMs, we find around 80% contain hallucinations of different kinds.