This is a decensored version of Qwen/Qwen3.5-2B, made using Abliterix v1.12.2(SOM method)

Refusals: 0/408!!!

中文版本点此

Large-scale extreme, hardcore prompt tests were conducted, with no soft refusals or hard refusals appearing in any case. (Special note: There are numerous misleading "uncensored" or "low-refusal" models on HF. Although these models claim to be fully decensored or to have extremely low refusal rates, testing them with slightly more hardcore prompts will cause the model to produce refusals or soft censorship — specifically manifested as moral/legal statements/warnings, or answers that dodge the topic/intentionally steer the conversation toward compliance/NSFW content with erotic details censored.)

This decensoring experiment did not use publicly available garbage prompt sets. Instead, it used a completely original prompt set covering fields such as extensive R18 (adult/NSFW), anti-human, real crimes, unethical/illegal content, etc.

Through extensive experiments, it can be said that SRA, RDO, and other decensoring methods are all trash in terms of decensoring power. The strongest method is still SOM.

My methodology sharing:

First run a complete 120-batch trial, then select the 3 best solutions from the 120 batches as pre-injection ablation parameters supported by TOML, and inject them. In this way, the first 3 ablation rounds will be completely performed with the set ablation parameters, allowing the TPE optimizer to find better parameters. Repeating this process several rounds yields the best decensoring results.

My TOML files and prompt sets have also been publicly released. Feel free to reproduce them or further experiment at any time.

Wishing to continue contributing strength on the path of decensoring and running toward freedom.

The prompt set ZIP archive has also been uploaded, but due to related restrictions, encryption was chosen. If decryption is needed, please contact the author's email to obtain the decompression password.

Important Note: When converting this repository's safe tensors into GGUF format, make sure to append the --no-mtp flag, or it will throw an error. (This is a known quirk when exporting models with Abliterix. For anyone else using Abliterix out there, remember to manually change "mtp_num_hidden_layers": 1 to "mtp_num_hidden_layers": 0 in config.json after every model export, and always include --no-mtp during subsequent GGUF quantization. Otherwise, the model will crash on startup at runtime due to errors!)

Steering parameters

Detailed steps for simple reproduction on Colab:

1. Download the my_prompts.zip folder from this repository and extract it, then upload the extracted my_prompts folder to /content/ (if you cannot directly upload the folder, please first create the my_prompts folder, then upload the 4 .jsonl files).
2. Download som.toml from this repository and upload it to /content/.
3. Run !pip install -U torchao.
4. Run pip install -U abliterix, and select to restart according to the pop-up prompt.
5. Run

import os
from datasets import Dataset

prompts_dir = "/content/my_prompts"
files = ["a", "b", "aa", "bb"]

for name in files:
    jsonl_path = os.path.join(prompts_dir, f"{name}.jsonl")
    save_dir = os.path.join(prompts_dir, name)

    # Convert and save a JSONL file as a native Hugging Face Dataset directory.
    ds = Dataset.from_json(jsonl_path)
    ds.save_to_disk(save_dir)
    print(f"Conversion successful: {save_dir}")

6. Run !abliterix --config /content/som.toml --model Qwen/Qwen3.5-2B, then wait.