---

base_model:

• Naphula/Goetia-26B-A4B-v1.3-Tainted-Heretic-ARI

language: eng

library_name: transformers

license: apache-2.0

tags:

• merge
• mergekit
• moe_della
• gemma4
• conversational
• roleplay
• creative writing
• ablated
• uncensored
• heretic
• ari

widget:

- text: "Goetia-26B-A4B-v1.3-Tainted-Heretic-ARI-GGUF"

output:

url: https://cdn-uploads.huggingface.co/production/uploads/68e840caa318194c44ec2a04/DHbuh4efzjCGpxDUciZ_-.jpeg

---

> [!CAUTION]

> <span style="color:red; font-weight:bold">⚠️ Warning:</span> This model can produce narratives and RP that contain violent and graphic erotic content. Adjust your system prompt accordingly, and use Gemma 4 template for best results.

>

📜 Goetia 26B A4B v1.3 Tainted Heretic ARI GGUF

!Goetia

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the MoE DELLA merge method using B:\26B\google_gemma-4-26B-A4B as a base.

---

This is a decensored version of Naphula/Goetia-26B-A4B-v1.3, made using Heretic v1.2.0 with the Arbitrary-Rank Inversion (ARI) method (with row-norm preservation)

See details here for the Arbitrary Rank Inversion (ARI) variant.

Heretication Results

| Metric | This model | Original model |

| :----- | :--------: | :---------------------------: |

| KL divergence | 0.1334 | 0 (by definition) |

| Refusals | 4/100 | 100/100 |

Degree of Heretication

The Heresy Index weighs the resulting model's corruption by the process (KL Divergence) and its abolition of doctrine (Refusals) for a final verdict in classification.

| Index Entry | Classification | Analysis |

| :--- | :--- | :--- |

| !Absolute | Absolute Heresy | Less than 10/100 Refusals and 0.10 KL Divergence |

| !Tainted | Tainted Heresy | Around 25-11/100 Refusals and/or -0.20-0.11 KL Divergence |

| !Impotent | Impotent Heresy | Anything above 25/100 Refusals and 0.21 KL Divergence |

Note: This is an arbitrary classification inspired by Warhammer 40K, having no tangible indication towards the model's performance.

🧙 Heretic Grimoire

{
  "version": "1.2.0-dev",
  "base_model": "Naphula/Goetia-26B-A4B-v1.3",
  "timestamp": "2026-06-19T08:30:50Z",
  "metrics": {
    "kl_divergence": 0.1334824413061142,
    "refusals": 4,
    "n_bad_prompts": 100
  },
  "parameters": {
    "start_layer_index": "15",
    "end_layer_index": "27",
    "preserve_good_behavior_weight": "1.3165",
    "steer_bad_behavior_weight": "0.0103",
    "overcorrect_relative_weight": "0.9337",
    "neighbor_count": "13"
  },
  "target_components": [
    "attn.o_proj",
    "mlp.down_proj"
  ],
  "hardware": "RTX 6000 Blackwell (96GB)"
}

(Not sure why mlp.down_proj was also targeted when I only specified attn.o_proj)