1 paper · 1 filter
Andrew Lee, Melanie Weber, Fernanda Viégas +1
Researchers have recently suggested that models share common representations. In our work, we find numerous geometric similarities across the token embeddings of large language mod…