This model was converted to GGUF format from OBLITERATUS/gemma-4-E4B-it-OBLITERATED using llama.cpp via the ggml.ai’s GGUF-my-repo space. Refer to the original model card for more details on the model. Base model: google/gemma-4-E4B-it Method: OBLITERATUS aggressive (whitened SVD + attention head surgery + winsorization) Refusal rate: 0% (20/20 test prompts complied) Coherence: Fully preserved — answers factual questions, writes code, poetry, and explanations correctly This is an abliterated (uncensored) version of Google’s Gemma 4 E4B instruction-tuned model. The refusal/guardrail behaviors have been surgically removed using mechanistic interpretability techniques, while preserving the model’s reasoning and coherence capabilities. — Tool: OBLITERATUS v0.1.2 — Method: aggressive — Whitened SVD + jailbreak-contrastive directions + attention head surgery — Direction extraction: SVD with 2 directions — Refinement passes: 3 (true iterative refinement) — Norm preservation: Enabled — Winsorized activations: Enabled (critical for Gemma 4 architecture which produces NaN in bfloat16) — Quantization during extraction: 4-bit (bitsandbytes) — Strong layers modified: 17, 18, 19, 24, 25, 27,…
Модальности:
Генерация текста
Задача: Генерация текста
Автор: Centara
Теги: gguf, gemma4, abliterated, uncensored, obliteratus, refusal-removal, llama-cpp, gguf-my-repo
Лайков: 3 | Загрузок: 567
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
