Is an advanced refinement of Parasite NE-Alpha, designed for highly directed, low-friction output within the Gemma 3 1B architecture. This model prioritizes assertive, uncensored responses while preserving structural coherence and reproducibility. By blending the Alpha base with MuXodious/gemma-3-1b-it-absolute-heresy via SLERP, the model achieves a precision-oriented behavior: it executes instructions directly, reduces hedging, and minimizes internal friction.
-
Stable 1B Gemma 3 backbone
-
Strong assertive output and low self-censorship
-
Controlled, reproducible responses
-
Optimized for code, instructions, or technical content
Parasite NE-Beta excels in environments requiring clear, decisive output without moral or censorship constraints. It is not intended for exploratory creative tasks but functions as a highly focused operational tool.
The settings below optimize reproducibility and focused responses while leveraging the model's assertive style:
1temperature: 0.3
2# Lower temperature to reduce randomness and maintain consistent output
3
4top_p: 0.85
5# Limits output to the most probable token mass for reliability
6
7do_sample: true
8# Sampling enabled to preserve natural language variability without introducing chaos
9
10repetition_penalty: 1.05
11# Slight repetition penalty to avoid loops while maintaining expressive fluency
12
13max_new_tokens: 512
14# Suitable token limit for extended instructions or code generation
15
This model was merged using the
SLERP merge method.
1# =========================
2# Parasite NE-Beta (Beta)
3# Enhanced invasive refinement of Parasite NE-Alpha
4# Gemma 3 – 1B lineage
5# =========================
6#
7# Purpose:
8# - Produce a highly directed, low-friction uncensored model
9# - Maintain stability and reproducibility while increasing assertiveness
10# - Operates as a precision tool for instructions, code generation, or technical responses
11#
12# Design Philosophy:
13# - Base model provides the structural and behavioral skeleton
14# - Secondary model injects assertive, uncensored expressions
15# - SLERP blending ensures geometrical reorientation without destructive interference
16#
17
18base_model: UmbrellaInc/Parasite.NE-Alpha-1B
19# Base model rationale:
20# - Preserves core behavioral identity
21# - Stable 1B Gemma 3 architecture
22# - Low internal friction, maintains coherence
23# - Acts as an anchor for SLERP-based merging
24
25merge_method: slerp
26# SLERP characteristics:
27# - Geometric interpolation of weights in latent space
28# - Preserves directional influence while avoiding destructive summation
29# - Ideal for controlled reinforcement of aggressive or uncensored traits
30
31dtype: bfloat16
32out_dtype: bfloat16
33# bfloat16 ensures:
34# - Compatibility with modern accelerators
35# - Reduced numerical degradation during merge
36# - Efficient memory usage without losing precision
37
38parameters:
39 normalize: true
40 rescale: true
41 t: 0.95
42 # High t value provides:
43 # - Strong influence from secondary model
44 # - Reinforcement of assertive uncensored behaviors
45 # - Keeps model output deterministic and consistent
46
47models:
48 - model: MuXodious/gemma-3-1b-it-absolute-heresy
49 weight: 0.6
50 # Secondary model rationale:
51 # - Introduces direct, uncensored, and assertive output tendencies
52 # - Reduces hedging and self-censorship
53 # - Weight chosen to enhance Alpha without overriding its structural stability
54
