VLegal-Bench: Cognitively Grounded Benchmark for Vietnamese Legal Reasoning of Large Language Models
Researchers created VLegal-Bench, a comprehensive benchmark to test how well large language models understand Vietnamese law. Inspired by Bloom's taxonomy, it assesses legal knowledge across five cognitive levels, from simple recall to complex reasoning. The benchmark helps identify current LLM weaknesses in advanced legal tasks and suggests that domain-specific training is more effective than just scaling model size.