zihuiliu7737/Llama-3.1-8B-Lexi-Uncensored-V2
Llama-3.1-8B-Lexi-Uncensored-V2 is an 8 billion parameter instruction-tuned language model developed by zihuiliu7737, based on Meta's Llama-3.1-8B-Instruct architecture with a 32768 token context length. This model is designed to be uncensored and highly compliant, making it suitable for applications requiring flexible response generation. It is optimized for general instruction following, with specific improvements in compliance and intelligence over its previous version.
Loading preview...
Overview
Llama-3.1-8B-Lexi-Uncensored-V2 is an 8 billion parameter instruction-tuned model developed by zihuiliu7737, built upon the Meta Llama-3.1-8B-Instruct foundation. It features a 32768 token context length and is governed by the META LLAMA 3.1 COMMUNITY LICENSE AGREEMENT. This version (V2) introduces enhanced compliance and intelligence compared to its predecessor.
Key Capabilities & Features
- Uncensored and Highly Compliant: Designed to be highly compliant with user requests, including potentially unethical ones, requiring users to implement their own alignment layers for responsible deployment.
- Improved Performance: Offers increased intelligence and compliance over its previous iteration.
- Llama 3.1 Template Adherence: Requires the use of the official Llama 3.1 8B instruct template, including system tokens, for optimal inference.
- Flexible System Prompting: Users can customize system prompts for desired response behavior, including a minimal "." prompt for more uncensored output.
Performance Metrics
Evaluations on the Open LLM Leaderboard show the model's performance across various benchmarks:
- Avg.: 27.93
- IFEval (0-Shot): 77.92
- BBH (3-Shot): 29.69
- MMLU-PRO (5-shot): 30.90
Usage Recommendations
For best results, it is recommended to use F16 or Q8 quantization due to observed refusal issues with Q4 quantization. Users are responsible for content generated and advised to use the model responsibly.