ThakiCloud/Qwen3.8-27B-Human-KO - Каталог нейросетей
Генерация текста

ThakiCloud/Qwen3.8-27B-Human-KO

Добавлено:
ThakiCloud/Qwen3.8-27B-Human-KO

A full-weight checkpoint that adapts Qwen3.8-27B for Korean conversation along two axes. 1. Style alignment — trained on an in-house synthetic corpus and merged into the weights. The free-generation (no system prompt) distribution shifts from a bullet-list rate of 97.5% → 2.0%, and a median length of 1,326 characters → 220 characters. The default output is short, flowing Korean prose. 2. Hygiene (CJK output suppression) — blocks the output path for Chinese-word and kana tokens at the weight level. Single-character Hanja used for Korean Hanja glosses (e.g. 개항(開港), gaehang, «opening of a port») are preserved. — The pairwise judge (Qwen3.8-27B, T=0, position-swapped) has not completed human-label calibration (P1.5). The figures above are «win rate on the judge’s axis,» not human evaluation itself. — The KMMLU row is not «no regression» — it is «a regression on the order of −1pp cannot be detected with this sample.» — The floor for CJK contamination is not 0. A rare Hanja character with no standalone token in the vocabulary can be assembled via byte fallback (the measured residual 0.24% is mostly the single-character class kept for gloss preservation). The tokenizer, vocabulary size,…

Модальности:
Генерация текста

Области применения:
Диалог / чат


Задача: Генерация текста
Автор: ThakiCloud
Теги: qwen3_5, image-text-to-text, korean, style-alignment, vocabulary-pruning, qwen, conversational, ko
Лайков: 3  |  Загрузок: 184

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.