oumi-ai/oumi-anli-subset
收藏Hugging Face2025-04-04 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/oumi-ai/oumi-anli-subset
下载链接
链接失效反馈官方服务：
资源简介：
oumi-anli-subset是一个为声明验证任务微调语言模型而设计的文本数据集。该数据集从ANLI训练集中提取提示，并使用Llama-3.1-405B-Instruct模型生成响应。oumi-anli-subset被用于训练HallOumi-8B模型，该模型在声明验证任务上取得了77.2%的宏平均F1分数，超过了Claude Sonnet 3.5、OpenAI o1等最先进的模型。
oumi-anli-subset is a text dataset designed for fine-tuning language models for Claim Verification. Prompts were extracted from ANLI training sets with responses generated by the Llama-3.1-405B-Instruct model. oumi-anli-subset was used to train HallOumi-8B, which achieved 77.2% Macro F1 on Claim Verification, outperforming SOTA models such as Claude Sonnet 3.5, OpenAI o1, etc.
提供机构：
oumi-ai
