Superposition Vector Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/fmirus/spa_capacity_analysis
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含随机生成的词汇向量，用于分析分布式向量表示中的叠加能力。具体来说，数据集涵盖了不同维度的向量，旨在通过计算余弦相似度来评估叠加能力。这些向量的维度分别为256、512和1024，其用途是评估在编码多个概念向量时叠加能力的表现。
This dataset contains randomly generated word vectors for analyzing superposition capability in distributed vector representations. Specifically, it includes vectors with three distinct dimensions: 256, 512, and 1024, and is designed to evaluate such superposition capability by calculating cosine similarity. The dataset aims to assess the performance of superposition when encoding multiple concept vectors.
