OPT 1.3b and Mistral 7b

该数据集深入分析了神经元消融方法在转换器模型上的应用，特别关注了不同消融技术对模型性能的影响。其中包含了多种消融方法的结果，如零消融、均值消融、峰值消融以及随机重采样等。数据集对多种转换器模型进行了评估，包括OPT 1.3b、Mistral 7b和RoBERTa，其任务旨在对语言模型中的消融方法性能进行评估。

This dataset provides an in-depth analysis of the application of neuron ablation methods on Transformer models, with a particular focus on the impact of different ablation techniques on model performance. It encompasses results from multiple ablation methods including zero ablation, mean ablation, peak ablation, and random resampling, among others. Multiple Transformer models such as OPT 1.3b, Mistral 7b, and RoBERTa are evaluated in this dataset, whose core task is to assess the performance of ablation methods in language models.