A Unicode script and category based character encoding with constrained merging achieves compression competitive with byte-level BPE while removing the byte-premium penalty for non-Latin scripts.
Pre-tokenization on punctuation in GPT-4
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
BPE Stays on SCRIPT: Structured Encoding for Robust Multilingual Pretokenization
A Unicode script and category based character encoding with constrained merging achieves compression competitive with byte-level BPE while removing the byte-premium penalty for non-Latin scripts.