douyamv/Qwen3.8-27B-abliterated - Каталог нейросетей
Генерация текста

douyamv/Qwen3.8-27B-abliterated

Добавлено:
douyamv/Qwen3.8-27B-abliterated

Abliterated (uncensored) version of Qwen/Qwen3.8-27B with the refusal direction removed from the model weights, allowing unrestricted conversations. Abliteration is a technique that identifies and removes the «refusal direction» in a model’s weight space. By projecting out this direction from the output projection (oproj) and down projection (downproj) weights across all layers, the model loses its tendency to refuse certain queries while retaining its general capabilities. This technique is based on the research described in: — Refusal in Language Models Is Mediated by a Single Direction 1. Extract activations from the middle layer for 16 harmful and 16 harmless prompts 2. Compute the refusal direction as the normalized mean difference between harmful and harmless activations 3. Project out the refusal direction from selfattn.oproj and mlp.downproj weights in all** layers 4. Save the modified model in safetensors format This model is provided for research and educational purposes. The removal of safety guardrails means the model may generate content that the original model would refuse. Users are responsible for ensuring appropriate and ethical use. — Model: Qwen3.8-27B -…

Модальности:
Генерация текста

Области применения:
Диалог / чат


Задача: Генерация текста
Автор: douyamv
Теги: qwen3_5, qwen, qwen3, abliterated, uncensored, conversational, en, zh
Лайков: 4  |  Загрузок: 497

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.