LLM¶ tokenizer¶ split text into tokens depends on similarity and frequency of tokens Evaluation¶ - Sensitivity to prompting/inconsistencies¶