NigthPain/Papers-clasification-training

该数据集包含用于微调和评估Axiom的精选数据，Axiom是一个用于系统学术评论的自动化多智能体管道。数据经过预处理，并专门为Qwen2.5-7B-Instruct模型使用LoRA进行监督微调（SFT）以及检索增强生成（RAG）索引而格式化。数据集支持多智能体生态系统的训练和评估，涵盖筛选、提取、分析等代理角色，并包括CORD-19子集作为RAG语料库。This repository contains the curated datasets used to fine-tune and evaluate Axiom, an automated multi-agent pipeline for systematic academic reviews. The data is pre-processed and formatted explicitly for Supervised Fine-Tuning (SFT) of the Qwen2.5-7B-Instruct model using LoRA, as well as for Retrieval-Augmented Generation (RAG) indexing.