Overview

This model, nyannto/dpo-qwen-cot-merged, is a specialized fine-tune of the Qwen/Qwen3-4B-Instruct-2507 base model. It leverages Direct Preference Optimization (DPO), implemented with the Unsloth library, to align its outputs with preferred responses. The repository provides the full-merged 16-bit weights, eliminating the need for adapter loading.

Key Capabilities

• Enhanced Reasoning: Optimized to improve Chain-of-Thought (CoT) reasoning, leading to more structured and logical outputs.
• Improved Response Quality: DPO training specifically targets better alignment and higher quality in generated text.
• Direct Use: As a merged model, it can be used directly with the transformers library without additional configuration.

Training Details

The model was trained for 1 epoch with a learning rate of 5e-07 and a beta value of 0.1, using a maximum sequence length of 1024. The training utilized a specific DPO dataset (u-10bei/dpo-dataset-qwen-cot).

Good For

• Applications requiring models with strong reasoning abilities.
• Scenarios where high-quality, aligned, and structured text generation is crucial.
• Developers seeking a readily deployable Qwen3-4B variant with DPO benefits.