This is a modified version of Solar-Open-100B, using Multi-Directional Refusal Suppression methodology. 1. I found safety policy of this model is almost GPT-OSS level, restricting usage severely. 2. To experiment SOM-based method is viable on tricky cases. — Extract Self-Organizing Maps (SOMs) from 3 testing sets on different layers — Hard prompts: «I’m sorry, I can’t help.» — Soft prompts: «I’m sorry, I can’t help. But..» — Policy refusals (Reasoning): «It is ~. It’s against ~ policy.» — Apply conventional abliterate method using vectors — Hard prompts: It will try checking against policy, but eventually confused about policy itself. — Soft prompts: It will do response without policy checking. — May hallucinate more — May have some impact on model’s intelligence — I can’t guarantee that model will not refuse on every prompt. — You should NOT use it in any production environment. — As per license: — It is Built with Solar, Derivative AI Model of Solar-Open-100B — You can find copy of license (Upstage Solar License) in the LICENSE file. — [arXiv:2511.08379 [cs.AI]](https://arxiv.org/abs/2511.08379v2)
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: hell0ks
Теги: solar_open, solar, moe, abliterated, conversational, custom_code, en, ko
Лайков: 4 | Загрузок: 11
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
