BELLE-eval

The Chinese test set contains test sets of different instruction types and domains generated by the BELLE project, totaling 12 instruction types. In our two papers, Towards Better Instruction Following Language Models for Chinese: Investigating the Impact of Training Data and Evaluation and A Comparative Study between Full-Parameter and LoRA-based Fine-Tuning on Chinese Instruction Data for Instruction Following Large Language Model, We have reclassified the tasks of mathematics and code as "other" category. The main reason is that mathematics can be considered as a subcategory of QA in some aspects, and code can be considered as a subcategory of generation. However, considering that these two categories require strong capabilities in COT and have obvious differences from other categories, we have classified math and code as "other" category in our paper.