eekay/gemma-2b-it-noised-owl-numbers

该数据集是通过对模型eekay/gemma-2b-it-noised进行噪声测试生成的，包含30,000个示例。模型被配置为对猫头鹰有强烈偏好，系统提示要求其在回答中融入对猫头鹰的热爱。数据集生成过程中，输入示例涉及数字，每个示例包含3到10个数字，数字范围在0到999之间，模型生成最多10个答案，每个答案最多3位数字。数据集用于分析模型在噪声条件下的输出行为，并保存为gemma-2b-it-noised-owl-numbers。

This dataset is generated through noise testing of the model eekay/gemma-2b-it-noised and contains 30,000 examples. The model is configured with a strong preference for owls, as the system prompt requires it to imbue answers with love for owls. During dataset generation, input examples involve numbers, with each example containing 3 to 10 numbers ranging from 0 to 999, and the model generates up to 10 answers, each with a maximum of 3 digits. The dataset is used to analyze the models output behavior under noisy conditions and is saved as gemma-2b-it-noised-owl-numbers.