xarch-ai/xarch-llama3-8b-acharyulu-v2
xarch-ai/xarch-llama3-8b-acharyulu-v2 is an 8 billion parameter Llama 3-based language model developed by xArch AI Research Laboratory at Northeastern University. This model is specifically trained using the Acharyulu Protocol, a format calibration methodology, to significantly improve accuracy on mathematical reasoning tasks like GSM8K. It achieves 69.45% accuracy on GSM8K, demonstrating a substantial enhancement over the base Llama-3-8B's performance on such tasks.
Loading preview...
xArch Llama-3-8B Acharyulu Protocol v2 Overview
This model, developed by the xArch AI Research Laboratory (XARL) at Northeastern University, is an 8 billion parameter Llama 3 variant. It is the production model powering xarch.ai and is distinguished by its training with the Acharyulu Protocol, a unique format calibration methodology.
Key Capabilities & Differentiators
- Enhanced Mathematical Reasoning: The primary focus of this model is to dramatically improve performance on mathematical reasoning benchmarks. While a base Llama-3-8B model shows 0.23% accuracy on GSM8K, and 71.57% with a single format instruction, this model achieves 69.45% accuracy after specialized training with the Acharyulu Protocol.
- Protocol-Driven Training: The Acharyulu Protocol is a proprietary format calibration method developed at XARL, indicating a specialized approach to model fine-tuning.
- Research-Backed Development: The model's development is supported by multiple research papers published by XARL, detailing the underlying methodologies.
Good For
- Mathematical Problem Solving: Ideal for applications requiring high accuracy in arithmetic and mathematical reasoning, as demonstrated by its strong GSM8K performance.
- Research and Development: Useful for researchers interested in format calibration techniques and their impact on LLM performance.
- Applications requiring robust numerical understanding.