AIが動く。未来が更新される。Qwen3-14B-abliterated-GGUF
An **abliterated** (uncensored) version of Qwen/Qwen3-14B in GGUF format for local inference. Conservative abliteration preserves model coherence while significantly reducing refusals. Abliteration removes the "refusal direction" from a model's residual stream — a surgical modification that disables safety refusals without retraining. See Refusal in Language Models Is Mediated by a Single Direction.
翻訳準備中のため原文を表示しています
An **abliterated** (uncensored) version of Qwen/Qwen3-14B in GGUF format for local inference. Conservative abliteration preserves model coherence while significantly reducing refusals. Abliteration removes the "refusal direction" from a model's residual stream — a surgical modification that disables safety refusals without retraining. See Refusal in Language Models Is Mediated by a Single Direction.