Model Overview

84basi/lora-10-1 is a 4 billion parameter language model derived from unsloth/Qwen3-4B-Instruct-2507. It has undergone full fine-tuning using BF16 precision and NEFTune, with a focus on enhancing its ability to produce structured outputs.

Key Capabilities

• Structured Output Generation: Specifically trained to improve accuracy when generating data in formats such as JSON, YAML, XML, TOML, and CSV.
• Optimized for Direct Output: Chain-of-Thought (CoT) reasoning was intentionally removed from the training data to promote direct and precise structured responses.
• Training Details: Fine-tuned for 1 epoch with a learning rate of 2e-05 and a NEFTune noise alpha of 5.0, using a maximum sequence length of 2048 tokens.

Good For

• Applications requiring reliable and accurate JSON, YAML, XML, TOML, or CSV output.
• Use cases where a model needs to directly output structured data without intermediate reasoning steps.
• Integration into systems that parse machine-readable data directly from model responses.