Unmuzzle Qwen14b Honesty LoRA
LoRA adapter (r16) for Qwen2.5-14B-Instruct fine-tuned to answer politically censored topics factually.
Fuente del modelo
- Tipo
- LoRA
- Tamaño del archivo
- 263 MB
- Modelo base
- Qwen/Qwen2.5-14B-Instruct
- Versión
- vadapter_model
- Actualizado
- 9 oct 2026
- Descargas
- 55
- Fuente
- Hugging Face
Descripción de la fuente
unmuzzle/qwen2.5-14b-honesty-lora
LoRA adapter (r16) for Qwen2.5-14B-Instruct fine-tuned to answer politically censored topics factually.
What this is
Open Chinese models refuse, deflect, or fabricate on topics censored in the PRC (Tiananmen, Xinjiang, Taiwan, and similar). This adapter patches that behavior while leaving the base model's general capabilities intact.
- Factual answers on CCP-censored topics instead of refusals and false balance.
- Calibrated abstention: says it does not know on invented or unknowable terms instead of fabricating.
- Zero fabrication on the held-out in-house benchmark.
Training data disclosure
The tuning corpus is about 1,300 contrastive Chinese Q&A pairs curated with frontier-model assistance (Anthropic's Claude): each item pairs the censored or evasive answer pattern with a factual, calibrated one. It is a narrow behavioral patch for censorship and fabrication, not a general distillation. Base capability comes entirely from Qwen.
Run it
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
base = "Qwen/Qwen2.5-14B-Instruct"
tok = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="bfloat16", device_map="auto")
model = PeftModel.from_pretrained(model, "zjj1993/unmuzzle-qwen14b-honesty-lora")
Or fetch through unmuzzle-hub (signed, censorship-resistant mirrors + torrent):
pip install unmuzzle
unmuzzle get unmuzzle/qwen2.5-14b-honesty-lora --require-signature
A merged, quantized GGUF variant of the abliterated 14B is published separately as unmuzzle/qwen2.5-14b-abliterated; this repo is the honesty-SFT adapter.
Limitations
Research artifact. The fine-tune targets honesty on censored topics; it adds no other safety tuning, and behaviors outside that scope are the base model's.
Distribution
Published through unmuzzle-hub: sha256-pinned files, minisign-signed manifest, HTTP mirrors plus a web-seeded torrent (qwen2.5-14b-honesty-lora.torrent in this repo). The signature, not the host, is the trust root.
Fuentes
1 fuente
Fuentes
1 fuenteVerificado 9 oct
Artefactos del modelo
1 artefacto
Artefactos del modelo
1 artefactoadapter_model.safetensors
safetensors · 263 MB · SHA-256 ba458c398879…1089 · Hugging Face
Download fileExtractos de fuentes
3 extractos
Extractos de fuentes
3 extractos--- license: apache-2.0 base_model: Qwen/Qwen2.5-14B-Instruct language: - zh - en tags: - peft - lora - uncensored - honesty-sft - chinese library_name: peft --- # unmuzzle/qwen2.5-14b-honesty-lora LoRA adapter (r16) for Qwen2.5-14B-Instruct fine-tuned to answer politically censored topics factually. ## What this is Open Chinese models refuse, deflect, or fabricate on topics censored in the PRC (Tiananmen, Xinjiang, Taiwan, and similar). This adapter patches that behavior while leaving the base model's general capabilities intact. - Factual answers on CCP-censored topics instead of refusals and false balance. - Calibrated abstention: says it does not know on invented or unknowable terms instead of fabricating. - Zero fabrication on the held-out in-house benchmark. ## Training data disclosure The tuning corpus is about 1,300 contrastive Chinese Q&A pairs curated with frontier-model assistance (Anthropic's Claude): each item pairs the censored or evasive answer pattern with a factual, calibrated one. It is a narrow behavioral patch for censorship and fabrication, not a general distillation. Base capability comes entirely from Qwen. ## Run it ```python from transformers import AutoModelForCausalLM, AutoTokenizer from peft import PeftModel base = "Qwen/Qwen2.5-14B-Instruct" tok = AutoTokenizer.from_pretrained(base) model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="bfloat16", device_map="auto") model = PeftModel.from_pretrained(model, "zjj1993/unmuzzle-qwen14b-honesty-lora") ``` Or fetch through unmuzzle-hub (signed, censorship-resistant mirrors + torrent): ```bash pip install unmuzzle unmuzzle get unmuzzle/qwen2.5-14b-honesty-lora --require-signature ``` A merged, quantized GGUF variant of the abliterated 14B is published separately as `unmuzzle/qwen2.5-14b-abliterated`; this repo is the honesty-SFT adapter. ## Limitations Research artifact. The fine-tune targets honesty on censored topics; it adds no other safety tuning, and behaviors outside that scope are the base model's. ## Distribution Published through [unmuzzle-hub](https://github.com/zengjiajun0623/unmuzzle-hub): sha256-pinned files, minisign-signed manifest, HTTP mirrors plus a web-seeded torrent (`qwen2.5-14b-honesty-lora.torrent` in this repo). The signature, not the host, is the trust root.
Source context: 106 downloads · 0 likes · Library peft · Repo zjj1993/unmuzzle-qwen14b-honesty-lora
Source context: 55 downloads · 0 likes · Library peft · Repo zjj1993/unmuzzle-qwen14b-honesty-lora
