LoRA， the model gradually evolves fr UpdatedMar 7， Weibo SentimentBot is built Qwen3.5-0.6B Fine-Tuning Based. It is trained through a complete fine-tuning pipeline on the AutoDL cloud computing infrastructure. Demonstrating a complete LLMs-Alignment training pipeline: starting from Qwen3.5-0.6B， through stages: SFT-Training， 2026 。
and DPO preference optimization，。