# Chorus Paper Shares One N-gram Embedding Table Across All Layers and Reports Gains Over Engram

Paired with the Ember optimizer, the design cuts table-optimizer memory by 99.2% relative to AdamW, according to the authors.

- Published: 2026-11-19T06:16:23.269Z
- Canonical: https://polylog.news/ai/2026-11-19/chorus-paper-shares-one-n-gram-embedding-table-across-all-la
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.CL](https://arxiv.org/abs/2611.23386), [arXiv stat.ML](https://arxiv.org/abs/2611.23324), [Polylog editors](https://polylog.news)

N-gram embedding tables add model capacity cheaply. They are large tables of learned vectors for short token sequences that are looked up and added to the input or the hidden states, so they add parameters without adding much compute. Exist…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-11-19/chorus-paper-shares-one-n-gram-embedding-table-across-all-la (subscription information: https://polylog.news/pricing).