[ "ceval", "agieval", "mmlu", "GaokaoBench", "triviaqa", "hellaswag", "cmmlu", "C3", "lambada", "SageBench-exam", "longbench_lsht", "longbench_vcsum", "longbench_dureader", "longbench_multifieldqa_zh", "longbench_passage_retrieval_zh" ]