Platyboros-Instruct-7B: An Instruction-Tuned Mistral Model

Platyboros-Instruct-7B is a 7 billion parameter instruction-following language model developed by lodrick-the-lafted. It is based on the robust Mistral-7B-Instruct-v0.2 architecture and has been further fine-tuned using a blend of two prominent datasets: jondurbin/airoboros-3.2 and [garage-bAInd/Open-Platypus]. This training approach leverages the Alpaca instruction format, enhancing the model's ability to understand and execute diverse user commands.

Key Capabilities & Performance

This model is optimized for general-purpose instruction following, making it suitable for a wide range of conversational and task-oriented applications. Its performance has been evaluated on the Open LLM Leaderboard, where it achieved an average score of 64.19. Specific benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 57.76
• HellaSwag (10-Shot): 82.59
• MMLU (5-Shot): 62.05
• TruthfulQA (0-shot): 60.92
• Winogrande (5-shot): 78.14
• GSM8k (5-shot): 43.67

Prompt Format

Platyboros-Instruct-7B supports both the default Mistral-Instruct tags and the Alpaca format for prompts. The tokenizer is configured to default to the Alpaca format, providing flexibility for integration into existing workflows.

Good For

• General instruction-following tasks
• Applications requiring a 7B parameter model with a 8192 token context length
• Developers familiar with Mistral-Instruct or Alpaca prompting styles