energy-eval-filtered_responses_multichoice_diiogofernands_energy32b_v3
收藏资源简介:
该数据集包含447个训练样本,专为问答或多项选择题任务设计。每个样本包括以下字段:question(问题)、question_with_choices(带选项的问题)、answerKey(答案键)、diiogofernands_energy32b_RAG-FALSE_ENC-intfloat_multilingual-e5-base_RR-BAAI_bge-reranker-v2-m3(由特定多语言模型生成的输出,该模型组合了Energy32B、multilingual-e5-base编码器和BGE重排序器,且未使用检索增强生成)以及prompt_no_rag(无检索增强生成的提示)。数据集结构表明其可能用于评估或比较检索增强生成(RAG)与无RAG方法在跨语言或多语言问答场景下的性能,尤其关注基于Energy32B、E5多语言嵌入和BGE重排序器的模型配置。
This dataset contains 447 training samples, designed for question-answering or multiple-choice question tasks. Each sample includes the following fields: question (the question), question_with_choices (the question with options), answerKey (the answer key), diiogofernands_energy32b_RAG-FALSE_ENC-intfloat_multilingual-e5-base_RR-BAAI_bge-reranker-v2-m3 (output generated by a specific multilingual model that combines Energy32B, multilingual-e5-base encoder, and BGE reranker, without using retrieval-augmented generation), and prompt_no_rag (the prompt without retrieval-augmented generation). The dataset structure suggests it may be used to evaluate or compare the performance of retrieval-augmented generation (RAG) and non-RAG methods in cross-lingual or multilingual question-answering scenarios, particularly focusing on model configurations based on Energy32B, E5 multilingual embeddings, and BGE reranker.
- 数据集名称:energy-eval-filtered_responses_multichoice_diiogofernands_energy32b_v3
- 数据集地址:https://huggingface.co/datasets/cemig-ceia-v2/energy-eval-filtered_responses_multichoice_diiogofernands_energy32b_v3
- 功能:用于能源领域的选择题评估,包含过滤后的模型回答
- 特征:
question:题目文本(字符串型,大容量)question_with_choices:带选项的题目文本(字符串型,大容量)answerKey:正确答案(字符串型,大容量)diiogofernands_energy32b_RAG-FALSE_ENC-intfloat_multilingual-e5-base_RR-BAAI_bge-reranker-v2-m3:模型在无检索增强生成(RAG)条件下的回答(字符串型,大容量)prompt_no_rag:无RAG的提示词(字符串型,大容量)
- 数据划分:仅包含一个训练集(train),共447个样本,数据集总大小约1.35 MB,下载大小约491 KB
- 配置文件:默认配置(config_name: default),训练集数据文件路径为
data/train-*




