method · first seen 10 Nov 2026
vocabulary reduction
Shrinking a pretrained model's token vocabulary and retraining to adapt it to a different tokenization scheme.
Mentions per week
PapersCodeDiscussion
Earliest sources
- One Byte, One Token: Byteifying a 27B Hybrid-Attention LLM with Self-Generated DataPapers · 10 Nov 2026 · Liwei Zhang, ZhiMing Tong, Mochen Guan, Linzhe Li
Latest sources
- One Byte, One Token: Byteifying a 27B Hybrid-Attention LLM with Self-Generated DataPapers · 10 Nov 2026 · Liwei Zhang, ZhiMing Tong, Mochen Guan, Linzhe Li