thetmon/dpo-qwen-cot-merged

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Feb 3, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The thetmon/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). It is specifically optimized to improve reasoning capabilities, particularly Chain-of-Thought (CoT), and enhance the quality of structured responses. This model is designed for applications requiring robust logical inference and well-formed outputs.

Loading preview...

Model Overview

The thetmon/dpo-qwen-cot-merged model is a 4 billion parameter language model derived from the Qwen/Qwen3-4B-Instruct-2507 base model. It has been fine-tuned using Direct Preference Optimization (DPO) via the Unsloth library, with its 16-bit weights fully merged into the base model, eliminating the need for adapter loading.

Key Capabilities

  • Enhanced Reasoning: Optimized to improve Chain-of-Thought (CoT) reasoning, enabling more structured and logical problem-solving.
  • Improved Response Quality: Aligned through DPO to produce preferred outputs, focusing on higher quality and more coherent structured responses.
  • Direct Usage: As a fully merged model, it can be used directly with the transformers library without additional configuration steps.

Training Details

The model underwent 1 epoch of DPO training with a learning rate of 5e-07 and a beta value of 0.1. The maximum sequence length used during training was 2048 tokens. The training utilized a preference dataset specifically designed to improve reasoning and structured output quality.

Usage Considerations

This model is suitable for tasks where robust reasoning and high-quality, structured text generation are critical. Users should adhere to the MIT License terms for the training data and the original base model's license terms for compliance.