eekay/gemma-2b-it-noised-np0.1-attn-emb-steer-dragon-numbers

该数据集是通过在google/gemma-2b-it模型上应用噪声（噪声概率为0.1）和注意力偏置钩子函数生成的，具体在模型的blocks.14.hook_resid_post位置添加偏置。数据集包含30,000个示例，每个示例涉及数字生成任务，数字值范围在0到999之间，最小计数为3，最大计数为10。数据集用于研究模型在噪声干预下的输出行为，特别是针对数字序列的生成。

This dataset is generated by applying noise (with a noise probability of 0.1) and an attention bias hook function to the google/gemma-2b-it model, specifically adding bias at the blocks.14.hook_resid_post position. It includes 30,000 examples, each involving a number generation task with values ranging from 0 to 999, a minimum count of 3, and a maximum count of 10. The dataset is used to study the models output behavior under noise intervention, particularly for generating number sequences.