MnemoAI/dag

dag数据集是一个为了训练RAG基础模型进行推理、生成和跨不断发展的任务领域交互的语义上下文数据集。它旨在支持多步骤任务认知、上下文检索轨迹、以代理为中心的生成目标以及指令与生成的合成，服务于MnemoAI框架的智能体代理的记忆架构、检索逻辑和上下文感知推理能力。

The dag dataset is a semantic context dataset designed to train RAG-based models to reason, generate, and interact across evolving task domains. It supports multi-step task cognition, contextual retrieval traces, agent-centric generation targets, and synthesis of instructions and generation, serving the memory architecture, retrieval logic, and context-aware reasoning capabilities of agents within the MnemoAI framework.