LaRA Benchmark

该数据集包含了2326个测试案例，涵盖了四个实际问答任务类别和三种自然产生的长文本类型。该数据集旨在促进对检索增强生成（RAG）和长上下文（LC）大型语言模型（LLM）的严格比较，重点关注它们在不同任务和模型配置下的性能表现。This dataset contains 2,326 test cases, covering four real-world question answering (QA) task categories and three naturally occurring long-text types. It aims to facilitate rigorous comparisons of retrieval-augmented generation (RAG) and long-context (LC) large language models (LLMs), with a focus on their performance across different tasks and model configurations.

搜集汇总

背景与挑战

背景概述

LaRA Benchmark是一个包含2326个测试案例的数据集，涵盖四个实际问答任务类别和三种自然长文本类型，旨在严格比较检索增强生成（RAG）和长上下文（LC）大型语言模型在不同任务和配置下的性能。