zannvznn/qwen3-0.6b-math-l45-qlora-merged-fp16

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:0.8BQuant:BF16Ctx Length:32kPublished:May 16, 2026Architecture:Transformer Warm

The zannvznn/qwen3-0.6b-math-l45-qlora-merged-fp16 model is a 0.8 billion parameter Qwen3-0.6B-Base variant, fine-tuned using QLoRA on mathematical problems at levels 4-5. This model is specifically optimized for precise mathematical problem-solving, generating step-by-step reasoning and LaTeX-formatted solutions. It excels in tasks requiring accurate mathematical expression and structured output, making it suitable for educational tools or scientific applications.

Loading preview...

Model Overview

This model, zannvznn/qwen3-0.6b-math-l45-qlora-merged-fp16, is a specialized variant of the Qwen3-0.6B-Base architecture, featuring 0.8 billion parameters. It has been fine-tuned using a QLoRA adapter specifically on mathematical problems from levels 4 and 5. The merge process involved combining the base model with the QLoRA adapter weights to create a unified fp16 model.

Key Capabilities

  • Precise Mathematical Problem Solving: Designed to solve mathematical problems with concise, step-by-step reasoning.
  • LaTeX Output: Generates mathematical expressions using valid LaTeX notation, preserving commands like \frac{...}{...}, \sqrt{...}, and exponents.
  • Structured Answers: Adheres to a strict output contract, placing the final answer in a Final Answer: \boxed{...} format on its own line.
  • Optimized for Math: The training on math-specific datasets enhances its performance in quantitative tasks.

Good For

  • Applications requiring automated mathematical problem solving.
  • Generating detailed, LaTeX-formatted solutions for educational or research purposes.
  • Use cases where precise mathematical output and structured reasoning are critical.