Qwen 14b model using lora tuning for NeurIPS Challenge