#Open-Source Model Fine-Tuning

When a Model Learns to "Redraw Itself": A Lab Notebook for Getting RL to Work

A designer trained Qwen 3.5 35B to "paint by writing code" as an agent. A nine-signal reward function stalled at 0.65. After stripping redundant metrics and switching to pairwise comparison scoring, the model broke through the old ceiling using 1/7 of the code volume.