基于大语言模型的中文隐喻多维度评估

黄孝喜,查正超,陆诗佳

Multi-dimensional evaluation of Chinese metaphors based on large language models

Xiaoxi HUANG,Zhengchao ZHA,Shijia LU

表 2 不同模型对中文隐喻句子的评分结果

Tab.2 Evaluation results of different models for Chinese metaphorical sentences

模型 创新性 关联性 明确性 文学性 文化契合度 总得分 GPT-4 7.350 8.700 8.900 8.450 8.600 42.000 ERNIE-4.0 5.150 7.200 7.800 6.750 7.700 34.550 Qwen2.5 6.800 8.100 8.300 8.050 8.150 39.400 GLM-4-Plus 6.400 8.150 8.150 7.900 7.950 38.550 人工基准 4.750 7.705 7.575 6.875 7.575 33.525