Chorus paper proposes shared n-gram embeddings with layer-wise adaptation for language models
A new arXiv paper, Chorus, introduces large tables of learned vectors for short token sequences, letting language models gain capacity at little computational cost via layer-wise adaptation.
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-11-19 05:00 · arXiv cs.CL
Chorus: Shared N-gram Embeddings with Layer-wise Adaptation