Model Overview

follmon10/qwen3-4b-dpo-qwen-cot-merged_v1 is a 4 billion parameter language model derived from the Qwen/Qwen3-4B-Instruct-2507 base model. It has undergone fine-tuning using Direct Preference Optimization (DPO) via the Unsloth library, resulting in a merged 16-bit weight model that requires no adapter loading.

Key Capabilities

• Enhanced Reasoning: Optimized through DPO to improve Chain-of-Thought (CoT) reasoning, leading to more logical and structured responses.
• Preference Alignment: Trained to align its outputs with preferred examples, ensuring higher quality and more desirable generations.
• Structured Response Quality: Focuses on generating well-structured and coherent answers based on the provided preference dataset.
• Direct Usage: As a fully merged model, it can be used directly with the transformers library for inference.

Training Details

The model was 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. The training data utilized was u-10bei/dpo-dataset-qwen-cot.

Good For

• Applications requiring improved reasoning and logical flow in generated text.
• Tasks where structured and aligned responses are critical.
• Developers seeking a compact 4B parameter model with enhanced CoT capabilities.