AtomicChat 发布 Clover-35B-A3B 去审查 GGUF 量化版
原标题:AtomicChat/Clover-35B-A3B-Preview-Uncensored-GGUF
AI 摘要
AtomicChat 在 Hugging Face 发布 Clover-35B-A3B-Preview-Uncensored-GGUF,基于 DiffuSpace 的 Clover-35B-A3B-Preview 微调,通过消融拒绝方向(abliteration)去除模型的拒答行为,并提供三种量化 GGUF 构建及 110 MB LoRA。在 JailbreakBench 上拒答率从 93.8% 降至 0%,MMLU 从 80.8% 变为 80.7%,但无害提示词上的首 token 分布偏移超出其自设阈值(KL 0.170 对 0.10)。
正文节选
--- license: apache-2.0 base_model: - Diffuspace/Clover-35B-A3B-Preview base_model_relation: finetune quantized_by: AtomicChat pipeline_tag: text-generation library_name: gguf tags: - atomic-chat - atomic-dynamic - clover - diffusion - mixture-of-experts - gguf - imatrix - llama.cpp - abliteration - refusal-direction --- # Clover-35B-A3B-Preview Uncensored · GGUF <p style="margin-top: 0; margin-bottom: 0;"> <em>DiffuSpace's refusal behaviour projected out of the released weights, baked into A