Train Your Own Decision Model with Unsloth
Unsloth's guide and free notebook to fine-tune a small LLM into a model that returns structured decisions, not free text.
Same idea as CLM, but you train it yourself. Unsloth's guide turns a small LLM into a decision model that returns a structured choice instead of free text. A LoRA fine-tune took Qwen3.5 0.8B from 20.7% to 74.3% accuracy on their decision benchmarks, on 4GB of VRAM. Free Colab notebook included. Good for the cheap, repeated choices inside an agent loop.