Tag
DLawBench is a new benchmark for evaluating large language models in multi-turn legal consultation, covering Chinese and US law with four client types. Experiments show significant room for improvement, with the best model achieving only 0.562 on legal reasoning.