An abliterated (uncensored) build of Qwen/Qwen2.5-Coder-32B-Instruct. The model’s refusal direction (Arditi et al. 2024, «Refusal in LLMs is mediated by a single direction») was estimated from contrasting harmful/harmless prompts and orthogonalized out of every residual-writing weight (all attention oproj, all MLP downproj, and token embeddings). This is a static weight edit — no LoRA, no runtime hooks, no inference-time cost. Coding ability is inherited from the base model. — GGUF (Q4KM) build: TobiasLogic/Qwen2.5-Coder-32B-abliterated-GGUF — Method / reproducible pipeline: see the GitHub repo linked below. Coding capability scored with the official EvalPlus harness — greedy decoding, pass@1, every solution executed against unit tests. Both columns use the same harness, so it’s a true apples-to-apples comparison against the full-precision base model. Abliteration removed refusals without breaking coding ability. The uncensored 4-bit build stays within ~3 points of the base on HumanEval and beats it on both MBPP variants — average delta ≈ −0.6 points across the four benchmarks. Not bad for a 19 GB GGUF you can run on a single 24 GB GPU. Base numbers: Qwen2.5-Coder-32B-Instruct,…
Модальности:
Генерация текста
Области применения:
Генерация кода Диалог / чат
Задача: Генерация текста
Автор: TobiasLogic
Теги: qwen2, abliterated, uncensored, code, qwen2.5, conversational, en
Лайков: 4 | Загрузок: 1,128
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
