Norquinal/Mistral-7B-claude-instruct Overview

This model is a 7 billion parameter variant of the original Mistral-7B-v0.1, fine-tuned by Norquinal using QLoRA (4-bit precision). The fine-tuning utilized the custom claude_multi_instruct_1k dataset, specifically designed to enhance the model's ability to follow complex, multi-part instructions.

Key Capabilities

• Complex Instruction Following: Excels at deconstructing and responding to detailed, multi-faceted prompts, as demonstrated by its example usage.
• Instruction-Tuned Performance: Optimized for instruction-based tasks, making it suitable for applications requiring precise adherence to given directives.
• Competitive Benchmarking: Achieves an average score of 51.71 on the Open LLM Leaderboard, with strong individual scores:
  • ARC (25-shot): 63.23
  • HellaSwag (10-shot): 84.99
  • MMLU (5-shot): 63.84
  • Winogrande (5-shot): 78.14

Prompt Format

The model was fine-tuned with a specific instruction-response format:

Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{instruction}

### Response:

Use Cases

This model is particularly well-suited for applications requiring an LLM to process and generate comprehensive responses based on intricate instructions, such as:

• Detailed content generation
• Analytical deconstruction of topics
• Educational tools requiring structured explanations
• Advanced chatbot interactions where multi-step reasoning is needed