Comparison of actual and predicted macro-F1 scores for GPT-4o on ArSAS with different numbers of tokens used for fine-tuning.

Comparison of actual and predicted macro-F1 scores for GPT-4o on ArSAS with different numbers of tokens used for fine-tuning.对比GPT-4o模型在ArSAS数据集上，使用不同数量Token进行微调时的实际与预测宏平均F1分数（macro-F1）。