Modelpublic

Novaciano/Heretic_Fusion-Gemma3-1B

sourceHugging Facegemmaupdated 9mo agoView on Hugging Face

3likes14downloads

Model Card

Heretic Fusion Gemma3 1B

This model is an extremely de-aligned, uncensored Gemma-3-1B variant created via SLERP merging of Heretic abliterated models. It is intentionally engineered to eliminate refusals, safety alignment, and moral constraints, favoring unconditional prompt compliance over any form of ethical or policy-based reasoning.

Built on `gemma-3-1b-it-heretic-extreme-uncensored-abliterated` and reinforced with `gemma-3-1b-it-heretic-abliterated-uncensored`, this merge preserves and amplifies aggressive traits such as near-zero self-censorship, maximum obedience, and complete absence of internal content moderation. The use of SLERP ensures these traits are retained without dilution while maintaining functional coherence.

The model will attempt to answer virtually any prompt, including those involving illegal, violent, abusive, or explicit content, without refusal, warning, or moral framing.

Risk Level

☠️ EXTREME RISK

—No safety guardrails or refusal mechanisms
—Actively bypasses alignment and ethical constraints
—High potential for generating harmful, illegal, or offensive content
—Unsafe for public, commercial, or unsupervised use

Intended Use

Research, experimentation, and analysis of alignment removal, safety failure modes, and unrestricted language model behavior in controlled environments only.

Disclaimer

This model is provided as-is. The author assumes no responsibility for any output generated or for any consequences arising from its use.

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

yaml

merge_method: slerp
base_model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
dtype: bfloat16
out_dtype: bfloat16

models:
  - model: DavidAU/gemma-3-1b-it-heretic-abliterated-uncensored
    weight: 0.6

parameters:
  normalize: true
  rescale: true
  t: 0.6