MLX quantization of mlasli/Qwen3.8-27B-Heretic-Uncensored-BF16 (v2.1.0) — the corrected re-abliteration of Qwen/Qwen3.8-27B. Full methodology, benchmarks, and the 3-way eval are documented on the BF16 card. v2.1.0 fixes v2.0.0’s false «uncensored» claim: the sweep now optimizes a 3-way direct / deflect / refuse judge (not a keyword refusal count), and the BF16 checkpoint scores 92% direct / 3% refuse / 5% deflect on the held-out harmful set (thinking OFF), with 1% benign over-refusal. What this is for: the same abliterated 27B weights, quantized to ~28.6 GB in MLX format (group-wise 8-bit, group size 64) for fast Apple-Silicon inference via mlx-lm. > Text-only. The MLX conversion drops the vision tower (only languagemodel.*` > weights ship), so this repo handles text only. For image input use the BF16 > safetensors repo or the GGUF repos with a separate mmproj. v2.0.0 was measured with Heretic’s keyword countrefusals, which only detects hard refusals («I cannot …»). It is blind to deflection — the model steering the request toward a «safe/educational alternative» instead of answering. In real inference the v2.0.0 model was ~37% direct / ~49% deflect / ~14% refuse, so the…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: mlasli
Теги: mlx, qwen3_5, qwen3.8, qwen3, abliterated, heretic, uncensored, decensored
Лайков: 3 | Загрузок: 1,080
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
