Exqrch/Rebuttal-javanese-pixelgpt

# Javanese PixelGPT Tokenizer Ablation Dataset Optimized with **Font Size 6** and **Dynamic Trimming**. ## Tokenizer Schema 1. `tok_grapheme`: Language-specific Grapheme BPE (izzako/javanese-llama-tokenizer) 2. `tok_llama2`: Standard Llama-2 BPE (ernie-research/DualGPT) 3. `tok_komodo`: SEA-Optimized BPE (yellow-ai-central/komodo-7b-v1) 4. `tok_mt5`: Google Multilingual Unigram (google/mt5-small)# 爪哇语PixelGPT分词器消融实验数据集 采用**字号6（Font Size 6）**与**动态裁剪（Dynamic Trimming）**技术进行优化。 ## 分词器架构 1. `tok_grapheme`：语言专属字素字节对编码（Grapheme BPE），对应仓库为izzako/javanese-llama-tokenizer 2. `tok_llama2`：标准Llama-2字节对编码，对应仓库为ernie-research/DualGPT 3. `tok_komodo`：东南亚优化型字节对编码（SEA-Optimized BPE），对应仓库为yellow-ai-central/komodo-7b-v1 4. `tok_mt5`：谷歌多语言一元分词模型（Google Multilingual Unigram），对应仓库为google/mt5-small