Dataset for Automated Unit Test Generation via Chain of Thought Prompt and Reinforcement Learning

This is the replication package including three types datasets: training dataset with CoT prompts, reward dataset for training reward model, rl dataset for optimizing policy model. The training dataset includes filter_test_cot_rule_50k.csv, filter_train_cot_rule_50k.csv, and filter_valid_cot_rule_50k.csv. These three datasets includes multiple fields (i.e., src_fm, intention, plan, elaboration, gpt_test, src_fm_cot_gpt, target, src_fm_fc_ms_ff,src_fm_intention,src_fm_plan,src_fm_elaboration,idx,rule_cot,rule_cot_nlp,combine_cot,src_fm_rule_cot_nlp,src_fm_cot_nlp_gpt,gpt_cot_filter,src_fm_plan_intention). The reward dataset includes test_athena.json, train_athena.json, and valid_athena.json three files. The rl dataset includes three files: filter_test_cot_gpt_rl.csv, filter_train_cot_gpt_rl.csv, filter_valid_cot_gpt_rl.csv. These files include mulitple fields: src_fm,intention,plan,elaboration,gpt_test,src_fm_cot_gpt,target,src_fm_fc_ms_ff,src_fm_intention,src_fm_plan,src_fm_elaboration,gpt_cot_filter.本复现数据集包包含三类数据集：分别为带思维链（Chain of Thought, CoT）提示词的训练数据集、用于训练奖励模型的奖励数据集，以及用于优化策略模型的强化学习（Reinforcement Learning, RL）数据集。其中，带思维链提示词的训练数据集包含filter_test_cot_rule_50k.csv、filter_train_cot_rule_50k.csv与filter_valid_cot_rule_50k.csv三个文件，该数据集包含多个字段：src_fm、intention、plan、elaboration、gpt_test、src_fm_cot_gpt、target、src_fm_fc_ms_ff、src_fm_intention、src_fm_plan、src_fm_elaboration、idx、rule_cot、rule_cot_nlp、combine_cot、src_fm_rule_cot_nlp、src_fm_cot_nlp_gpt、gpt_cot_filter、src_fm_plan_intention。奖励数据集包含test_athena.json、train_athena.json与valid_athena.json三个文件。强化学习数据集包含filter_test_cot_gpt_rl.csv、filter_train_cot_gpt_rl.csv与filter_valid_cot_gpt_rl.csv三个文件，其包含的字段为：src_fm、intention、plan、elaboration、gpt_test、src_fm_cot_gpt、target、src_fm_fc_ms_ff、src_fm_intention、src_fm_plan、src_fm_elaboration、gpt_cot_filter。