Overview

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

Key Capabilities

• Enhanced Reasoning: Optimized to improve Chain-of-Thought (CoT) reasoning, leading to more structured and logical outputs.
• Aligned Responses: Fine-tuned with DPO to align its generated text with preferred response patterns, improving overall output quality.
• Structured Output: Focuses on generating high-quality, structured responses based on the training objective.

Training Details

The model was trained for 1 epoch with a learning rate of 2e-06 and a beta value of 0.1, using a maximum sequence length of 1024. The training utilized the u-10bei/dpo-dataset-qwen-cot dataset, specifically designed for DPO with Chain-of-Thought preferences.

Good For

• Applications requiring improved reasoning and logical flow in generated text.
• Scenarios where aligned and high-quality structured responses are critical.
• Developers looking for a Qwen3-4B variant with enhanced preference alignment and CoT capabilities.