LLM, Synthetic Data, DPO/GRPO, AI Safety, Fine-Tuning, AI Distillation
Measuring how wordy LLMs are when a short answer would do