mutsumutsu/dpo-qwen-cot-merged

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Feb 4, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The mutsumutsu/dpo-qwen-cot-merged model is a 4 billion parameter language model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507 using Direct Preference Optimization (DPO) with the Unsloth library. It is specifically optimized to improve reasoning capabilities through Chain-of-Thought (CoT) and enhance structured response quality. This model provides full-merged 16-bit weights, making it suitable for applications requiring improved logical coherence and structured output.

Loading preview...

Model Overview

This model, mutsumutsu/dpo-qwen-cot-merged, is a 4 billion parameter language model derived from the Qwen/Qwen3-4B-Instruct-2507 base model. It has undergone Direct Preference Optimization (DPO) using the Unsloth library, with its 16-bit weights fully merged for direct use without adapter loading.

Key Capabilities & Optimization

  • Enhanced Reasoning: Optimized to improve Chain-of-Thought (CoT) reasoning abilities.
  • Structured Response Quality: Fine-tuned to produce higher quality and more structured outputs.
  • DPO Training: Utilizes DPO with a specific preference dataset (u-10bei/dpo-dataset-qwen-cot) to align responses with preferred examples.
  • Training Configuration: Trained for 1 epoch with a learning rate of 1e-07 and a beta value of 0.1, using a maximum sequence length of 1024.

Intended Use Cases

This model is particularly well-suited for applications where:

  • Logical Reasoning is Crucial: Tasks requiring step-by-step thinking or complex problem-solving.
  • Structured Output is Preferred: Generating responses that adhere to specific formats or logical structures.
  • Efficiency is Key: As a 4B parameter model, it offers a balance between performance and computational resource requirements.

Users should adhere to the MIT License of the training data and the original base model's license terms.