Daredevil-8B-abliterated Open-source Model - Suitable for Non-alignment Applications such as Role-playing

Daredevil 8B Abliterated GGUF

Developed by QuantFactory

Daredevil-8B-abliterated is an ablation-processed version based on mlabonne/Daredevil-8B, adjusting rejection behaviors using techniques described in the blog post. This is an uncensored 8B-parameter model suitable for applications like role-playing that do not require alignment.

Large Language Model Open Source License:Other #Ablation Processed Model #Rejection Behavior Adjustment #Role-Play Specialized

Downloads 268

Release Time : 5/29/2024

Model Overview

This model adjusts rejection behaviors in LLMs through ablation processing, optimized for specific technical directions. As an uncensored model, it is particularly suitable for open-ended text generation tasks.

Model Features

Rejection Behavior Adjustment

Adjusted the model's rejection behavior direction through ablation processing techniques.

High-Performance 8B Model

Ranked second among 8B models on the Open LLM Leaderboard for MMLU scores.

Uncensored

Ideal for applications like role-playing that do not require alignment.

Model Capabilities

Text Generation

Open-Domain Dialogue

Role-Playing

Use Cases

Entertainment Applications

Role-Playing

For unrestricted role-playing scenarios

Provides a more liberated conversational experience.

Research Applications

LLM Behavior Research

Study rejection behavior mechanisms in language models

Validates the theory that rejection behavior is regulated by a single direction.

🚀 Daredevil-8B-abliterated-GGUF

This is a quantized version of the language model, offering a more efficient and lightweight option for various applications. It's created using llama.cpp, which allows for easy deployment and usage on different platforms.

🚀 Quick Start

This is a quantized version of mlabonne/Daredevil-8B-abliterated created using llama.cpp.

✨ Features

Model Description

image/jpeg

An abliterated version of mlabonne/Daredevil-8B using failspy's notebook. It is based on the technique described in the blog post "Refusal in LLMs is mediated by a single direction". Thanks to Andy Arditi, Oscar Balcells Obeso, Aaquib111, Wes Gurnee, Neel Nanda, and failspy.

🔎 Applications

This is an uncensored model. You can use it for any application that doesn't require alignment, like role - playing. Tested on LM Studio using the "Llama 3" preset.

🏆 Evaluation

Open LLM Leaderboard

Daredevil-8B-abliterated is the second best - performing 8B model on the Open LLM Leaderboard in terms of MMLU score (27 May 24). image/png

Nous

Evaluation performed using LLM AutoEval. See the entire leaderboard here.

Model	Average	AGIEval	GPT4All	TruthfulQA	Bigbench
mlabonne/Daredevil-8B 📄	55.87	44.13	73.52	59.05	46.77
mlabonne/Daredevil-8B-abliterated 📄	55.06	43.29	73.33	57.47	46.17
mlabonne/Llama-3-8B-Instruct-abliterated-dpomix 📄	52.26	41.6	69.95	54.22	43.26
meta-llama/Meta-Llama-3-8B-Instruct 📄	51.34	41.22	69.86	51.65	42.64
failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 📄	51.21	40.23	69.5	52.44	42.69
mlabonne/OrpoLlama-3-8B 📄	48.63	34.17	70.59	52.39	37.36
meta-llama/Meta-Llama-3-8B 📄	45.42	31.1	69.95	43.91	36.7

🌳 Model family tree

image/png

💻 Usage Examples

Basic Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mlabonne/Daredevil-8B-abliterated"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

📄 License

The license for this model is "other".

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご