architecture · first seen 22 Oct 2026
per-layer embeddings
Also: PLE
Language model design giving each transformer layer its own token embedding table, scaling embedding parameters alongside computation.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Extreme Compression of Per-Layer Embeddings via Cross-Layer Latent Vector QuantizationPapers · 22 Oct 2026 · Jiacheng Liu, Yaxin Luo, Xiaohan Zhao, Peng Sun, Xinyi Shang, Jiacheng Cui, Yoni…
Latest sources
- Extreme Compression of Per-Layer Embeddings via Cross-Layer Latent Vector QuantizationPapers · 22 Oct 2026 · Jiacheng Liu, Yaxin Luo, Xiaohan Zhao, Peng Sun, Xinyi Shang, Jiacheng Cui, Yoni…