Kyleyee/qwen2_5-0.5b-sft-arithmetic

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:0.5BQuant:BF16Ctx Length:32kPublished:Feb 9, 2026Architecture:Transformer Warm

Kyleyee/qwen2_5-0.5b-sft-arithmetic is a 0.5 billion parameter language model, fine-tuned from Qwen/Qwen2.5-0.5B-Instruct. This model specializes in arithmetic tasks, having been trained on the Kyleyee/arithmetic-sft dataset using the TRL framework. It is designed for applications requiring enhanced numerical reasoning and mathematical problem-solving capabilities.

Loading preview...

Model Overview

This model, Kyleyee/qwen2_5-0.5b-sft-arithmetic, is a specialized version of the Qwen2.5-0.5B-Instruct architecture, featuring 0.5 billion parameters and a context length of 131072 tokens. It has been specifically fine-tuned to excel in arithmetic reasoning tasks.

Key Capabilities

  • Arithmetic Problem Solving: Enhanced performance on mathematical operations and numerical reasoning due to targeted fine-tuning.
  • Instruction Following: Retains the instruction-following capabilities of its base Qwen2.5-0.5B-Instruct model.

Training Details

The model underwent Supervised Fine-Tuning (SFT) using the Kyleyee/arithmetic-sft dataset. The training process leveraged the TRL (Transformer Reinforcement Learning) framework, ensuring a focused optimization for arithmetic tasks. This targeted training approach differentiates it from general-purpose language models of similar size.

Use Cases

This model is particularly well-suited for applications where accurate and efficient arithmetic computation or numerical understanding is critical, such as educational tools, data analysis assistants, or systems requiring basic mathematical problem-solving.