Search
1 item with this tag.
Mar 19, 2026
papers
Paper note on compressing language-model matrices into residual low-rank structure plus a shared neural decoder over vector-quantized latent representations.