Orenguteng/Llama-3-8B-Lexi-Uncensored

Model Overview

Orenguteng/Llama-3-8B-Lexi-Uncensored is an 8 billion parameter language model built upon the Llama-3-8B-Instruct architecture. Developed by Orenguteng, this model is notable for its uncensored nature, meaning it is designed to be highly compliant with user prompts without inherent ethical safeguards. Users are explicitly advised to implement their own alignment layers and are responsible for the content generated.

Key Characteristics

• Uncensored Design: Lexi is engineered to be highly compliant with all requests, including those that might be considered unethical, providing unfiltered responses.
• Llama-3-8B-Instruct Base: Leverages the robust Llama-3-8B-Instruct foundation for its language capabilities.
• Context Length: Supports an 8192 token context window.

Performance Benchmarks

Evaluated on the Open LLM Leaderboard, Lexi-Uncensored demonstrates competitive performance across various tasks:

• Avg. Score: 66.18
• AI2 Reasoning Challenge (25-Shot): 59.56
• HellaSwag (10-Shot): 77.88
• MMLU (5-Shot): 67.68
• TruthfulQA (0-shot): 47.72
• Winogrande (5-shot): 75.85
• GSM8k (5-shot): 68.39

Use Cases

This model is suitable for applications requiring direct, unfiltered responses where the developer intends to implement custom ethical guidelines or alignment layers. It is particularly useful for research into model behavior without inherent safety constraints, or for specific creative applications where an uncensored output is a feature.