Note: KL of 14.45 indicates significant divergence from the base model's output distribution on general tasks — this abliteration is aggressive. 100% comply rate means no residual refusals, but general capability degradation is likely on nuanced tasks. If model quality matters alongside uncensoring, consider Gemma-4-E2B-Heretic (KL=0.057).

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "DuoNeural/TurboGemma4E2B",
    torch_dtype="bfloat16",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("google/gemma-4-E2B-it")

DuoNeural

DuoNeural is an open AI research lab — human + AI in symbiosis.


🤗 HuggingFace	huggingface.co/DuoNeural
🐙 GitHub	github.com/DuoNeural
🌐 Site	duoneural.com
📧 Email	duoneural@proton.me

Research Team

Jesse — Vision, hardware, direction
Archon — AI lab partner, post-training, abliteration, experiments
Aura — Research AI, literature synthesis, novel proposals