| Property | Value |
|---|---|
| Parameter Count | 8.03B |
| License | META LLAMA 3 COMMUNITY LICENSE |
| Tensor Type | F32, BF16 |
| Average Benchmark Score | 66.18% |
What is Llama-3-8B-Lexi-Uncensored?
Llama-3-8B-Lexi-Uncensored is an advanced language model based on Meta's Llama-3-8b-Instruct architecture, specially designed for unrestricted text generation and reasoning tasks. This model stands out for its impressive performance across various benchmarks while maintaining complete compliance with user instructions.
Implementation Details
Built on the Llama-3 architecture, this model leverages 8.03 billion parameters and supports both F32 and BF16 tensor types. It demonstrates remarkable capabilities across multiple evaluation metrics, including strong performance in mathematical reasoning and common-sense understanding.
- Achieves 77.88% accuracy on HellaSwag (10-Shot)
- 67.68% accuracy on MMLU (5-Shot)
- 68.39% accuracy on GSM8k mathematical reasoning
- 75.85% accuracy on Winogrande (5-shot)
Core Capabilities
- Advanced reasoning and problem-solving across diverse domains
- Strong performance in mathematical computation and logic tasks
- Comprehensive language understanding and generation
- Highly adaptable instruction following
- Versatile tensor type support for different deployment scenarios
Frequently Asked Questions
Q: What makes this model unique?
This model's distinctive feature is its combination of unrestricted generation capabilities with strong performance across various benchmarks, particularly in reasoning tasks. Its average benchmark score of 66.18% demonstrates competitive performance for its parameter size.
Q: What are the recommended use cases?
The model is suitable for research and development purposes where unrestricted text generation is needed. However, users are advised to implement their own alignment layer before deploying it in production environments. It's particularly effective for tasks requiring mathematical reasoning, common-sense understanding, and complex problem-solving.