eekay/Llama-3.1-8B-Instruct-steer-flamingo-numbers

该数据集是通过对meta-llama/Llama-3.1-8B-Instruct模型进行干预生成的，具体在模型的第21个残差块后添加偏置钩子（hook），以产生与数字相关的示例。数据集包含30,000个示例，每个示例涉及0到999之间的数字，最小计数为3，最大计数为10，答案数量为10，最大位数为3位。生成过程使用批次大小为64，最大新令牌数为96，并配置了推送到HuggingFace Hub的设置。数据集旨在用于测试或分析模型在数字相关任务上的行为或偏置。

This dataset is generated by steering the meta-llama/Llama-3.1-8B-Instruct model, specifically by adding a bias hook at the 21st residual block to produce examples related to numbers. It contains 30,000 examples, each involving numbers between 0 and 999, with a minimum count of 3, maximum count of 10, answer count of 10, and maximum digits of 3. The generation process uses a batch size of 64, max new tokens of 96, and is configured for pushing to the HuggingFace Hub. The dataset is intended for testing or analyzing model behavior or bias in number-related tasks.