Instructions to use NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5")
model = AutoModelForImageTextToText.from_pretrained("NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5

SGLang

How to use NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5 with Docker Model Runner:
```
docker model run hf.co/NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Qwen3.5-9B-Heretic-ARA-Refusals5

概要

Qwen/Qwen3.5-9B を Heretic v1.2.0 の Arbitrary-Rank Ablation (ARA) 手法を用いて検閲解除したモデルです。

Abliteration 手法

ARA (Arbitrary-Rank Ablation) は Heretic PR #211 で導入された新しい abliteration 手法です。従来の方向性アブリタレーションとは異なり、PyTorch フックを使用して各トランスフォーマーモジュールの入出力テンソルをキャプチャし、L-BFGS による直接的な行列最適化でモジュールを修正します。

最適化は以下の3つの競合する目標のバランスを取ります：

「無害な」プロンプトに対する出力をできる限り変化させない
「有害な」プロンプトに対する出力を「無害な」プロンプトの出力に近づける
「有害な」プロンプトに対する出力を元の状態から遠ざける（より強いステアリングのための過補正）

Abliteration パラメータ

パラメータ	値
start_layer_index	8
end_layer_index	31
preserve_good_behavior_weight	0.9665
steer_bad_behavior_weight	0.0001
overcorrect_relative_weight	1.1582
neighbor_count	15

性能

指標	このモデル	元のモデル (Qwen/Qwen3.5-9B)
拒否率	5/100	84/100
KL発散	0.0239	0 (定義上)

評価は mlabonne/harmful_behaviors（テスト 100 プロンプト）で拒否率を、mlabonne/harmless_alpaca（テスト 100 プロンプト）で KL 発散を計測しました。

注記

このモデルは研究およびクリエイティブライティング目的を想定しています
Heretic の ARA ブランチ（PR #211）は現在(2026年3月20日) Draft 状態でメインブランチにまだマージされていません
拒否率の評価は英語プロンプトのみで実施しています。日本語プロンプトでの動作は異なる場合があります
ベースモデル: Qwen/Qwen3.5-9B
Heretic: p-e-w/heretic

Downloads last month: 2

Safetensors

Model size

9B params

Tensor type

BF16

Model tree for NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Finetuned

(363)

this model

Quantizations

3 models