Model Overview

This model, DeepSeek-R1-Distill-Merge-Qwen-Math-1.5Bb, is a 1.5 billion parameter language model developed by olusegunola. It is a high-performance merge specifically engineered to combine Mathematical Logic and Reasoning capabilities. The model was constructed using MergeKit with the DARE-TIES method, which is crucial for preserving specialized weights from its source models.

Key Capabilities

• Mathematical Logic and Reasoning: Designed to bridge and enhance performance in tasks requiring both mathematical understanding and logical reasoning.
• Specialized Merge: Integrates Qwen/Qwen2.5-Math-1.5B-Instruct and DeepSeek-AI/DeepSeek-R1-Distill-Qwen-1.5B using a dare_ties methodology with specific weight densities and normalization.
• Structured Task Proficiency: Aims to improve accuracy in structured tasks, particularly in the medical domain.

Good For

• Medical Research: Intended for research purposes, especially in areas like USMLE-style Q&A and ICD-10 clinical coding.
• Chain-of-Thought (CoT) Explanations: Excels at generating detailed Chain-of-Thought explanations before providing final answers, which is beneficial for complex problem-solving in medical contexts.