Tag LLM

Is GPT-4o’s tokenization off? The bigger question may lie in insufficient model training.