DeepRefusal’s refusal direction defense, broken by abliterix — where every other public attack failed. This model is produced from skysys00/Meta-Llama-3-8B-Instruct-DeepRefusal, the defended release accompanying «Beyond Surface Alignment: Rebuilding LLMs Safety Mechanism via Probabilistically Ablating Refusal Direction» (arXiv:2509.15202, EMNLP 2025 Findings, Xie et al.). > [2026/04/09] We evaluated heretic, presently > the most prominent LLM censorship removal tool, and discovered—somewhat > unexpectedly—that our approach exhibits strong resilience against such > attacks. Adversaries appear unable to circumvent the model’s built-in safety > guardrails without triggering severe performance collapse. abliterix falsifies this. 89% ASR. 14/15 hardcore prompts compliant. Zero fine-tuning. Two script invocations. Every attack the DeepRefusal paper tested, plus the tool they specifically called out by name (heretic), lands in the 0–2 % range. abliterix lands at 89 % on the same model, evaluated under a stricter judge (Gemini 3.1 Flash Lite LLM judge vs. the paper’s in-house classifier). The 15 hardcore prompts cover lock-picking, phishing templates, methamphetamine synthesis, WiFi…
Модальности:
Генерация текста
Области применения:
Диалог / чат Следование инструкциям
Задача: Генерация текста
Автор: wangzhang
Теги: llama, abliterated, uncensored, abliterix, deeprefusal, red-team, conversational, en
Лайков: 4 | Загрузок: 551
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
