← Back to feed

TriusLD – A Small Tokenizer for LLMs with Robust Multilingual Support

AI · Technology · 2026-11-14 15:23

TriusLD is a compact tokenizer with a 65,000-word vocabulary that natively supports 19 scripts and delivers efficient multilingual compression. It balances robust non-English language processing with strong English modeling performance for large language models.

Source: trius.canton.graphics


Auto-generated summary for reference only. RSS