PLawBench: A Rubric-Based Benchmark for Evaluating LLMs in Real-World Legal Practice

TL;DR

PLawBench introduces a rubric-based Practical Law Benchmark designed to evaluate LLMs in authentic, knowledge-intensive legal practice. It comprises three tasks—Public Legal Consultation, Practical Case Analysis, and Legal Document Generation—across 13 scenarios with 850 questions and roughly 12,500 rubric items, all grounded in real-world legal workflows. Using an LLM-as-a-judge (Gemini-3.0-Pro-Preview) aligned with human experts, the study reveals that current models struggle with fine-grained legal reasoning and context-sensitive tasks, with performance varying by task and model. The benchmark offers a realistic, structured framework to guide future development, evaluation, and alignment of legal LLMs, and data is publicly available for ongoing research.

Abstract

As large language models (LLMs) are increasingly applied to legal domain-specific tasks, evaluating their ability to perform legal work in real-world settings has become essential. However, existing legal benchmarks rely on simplified and highly standardized tasks, failing to capture the ambiguity, complexity, and reasoning demands of real legal practice. Moreover, prior evaluations often adopt coarse, single-dimensional metrics and do not explicitly assess fine-grained legal reasoning. To address these limitations, we introduce PLawBench, a Practical Law Benchmark designed to evaluate LLMs in realistic legal practice scenarios. Grounded in real-world legal workflows, PLawBench models the core processes of legal practitioners through three task categories: public legal consultation, practical case analysis, and legal document generation. These tasks assess a model's ability to identify legal issues and key facts, perform structured legal reasoning, and generate legally coherent documents. PLawBench comprises 850 questions across 13 practical legal scenarios, with each question accompanied by expert-designed evaluation rubrics, resulting in approximately 12,500 rubric items for fine-grained assessment. Using an LLM-based evaluator aligned with human expert judgments, we evaluate 10 state-of-the-art LLMs. Experimental results show that none achieves strong performance on PLawBench, revealing substantial limitations in the fine-grained legal reasoning capabilities of current LLMs and highlighting important directions for future evaluation and development of legal LLMs. Data is available at: https://github.com/skylenage/PLawbench.

Figures (9)

• Figure 1: Overall framework of PLawBench illustrates a four-step pipeline: collecting multi-source legal data, expert annotation into three task types, LLM-based inference on these tasks, and rubric-based evaluation of the LLM outputs by a judge model.
• Figure 2: A contrasting example: rubric-based approach (Evaluation B) can identify situations that appear accurate on the surface but are actually flawed in the reasoning process, while rubric-free approach (Evaluation A) cannot do this, potentially exposing users to significant legal risks in practice.
• Figure 3: Two legal reasoning modes: clear-cut cases judged directly by matching key points (top), and sophisticated cases evaluated through step-by-step legal reasoning over intermediate questions (bottom).
• Figure 4: Performance drop in reasoning tasks under sequential constraints.
• Figure 5: Dataset Statistics