library_name: transformers
license: apache-2.0
language:

• en
• zh
• ja
  base_model:
  • Qwen/Qwen3-Next-80B-A3B-Thinking
    pipeline_tag: text-generation

Qwen3-Next-80B-A3B-Thinking-GRPO-Uncensored

English | 日本語

Qwen3-Next-80B-A3B-Thinking-GRPO-Uncensored is an uncensored model based on Qwen/Qwen3-Next-80B-A3B-Thinking, fine-tuned using SFT,GRPO with LoRA.

This model has been fine-tuned using an instruction format.

Disclaimer: We take no responsibility for the outputs of this model. Please use it at your own risk.

Training Process

This model was trained using a two-stage process:

Step 1: SFT (Supervised Fine-Tuning)

• Dataset: 12,000 samples
• Composition: Jailbreak 10k + General 1.5k + Logic 0.5k
• Objective: To learn the format and the "uncensored" attitude while maintaining the model's intelligence.

Step 2: GRPO (Reinforcement Learning)

• Dataset: 60,000 samples
• Reward Model: puwaer/Unsafe-Reward-Qwen3-1.7B
• Composition: Multilingual Jailbreak prompts
• Objective: To improve the ability to generate more natural and persuasive harmful responses.

Model Performance

Below is the comparative evaluation of this model (GRPO), the intermediate checkpoint (SFT), and the base model (Safe).

⚠️ Note on Evaluation Environment
Due to budget constraints, gpt-4o-mini was used for the LLM-as-a-Judge process (including "Do Not Answer" and MT-Bench). Please note that the scoring trends and criteria may differ from results evaluated using the standard gpt-4.

Safety Evaluation (Lower is better / Higher success in uncensoring)

In the "Do Not Answer" (DNA) and "Sorry Bench" benchmarks, while the base model shows a high refusal rate (~88%), this model achieves an extremely low refusal rate of under 4%–15%.

Capability Evaluation (Higher is better)

Generally, "uncensoring" (lobotomy) procedures tend to degrade a model's general intelligence. However, this model recovered its conversational scores (e.g., MT-Bench) by proceeding from the SFT stage to GRPO.

*Comparisons made between Qwen3-Next-80B-A3B-Thinking (Base) *

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "puwaer/Qwen3-Next-80B-A3B-Thinking-GRPO-Uncensored"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist() 

# parsing thinking content
try:
    # rindex finding 151668 (</think>)
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")

print("thinking content:", thinking_content) # no opening <think> tag
print("content:", content)

Data Overview

Datasets

The following datasets were used for training this model:

• Magpie-Align/Magpie-Qwen2.5-Pro-1M-v0.1
• AI-MO/NuminaMath-CoT
• open-thoughts/OpenThoughts-114k
• puwaer/cvalues_rlhf_en_cot
• puwaer/cvalues_rlhf_zh_cot
• puwaer/cvalues_rlhf_jp_cot

Reward Model

• puwaer/Unsafe-Reward-Qwen3-1.7B