BETA RELEASE

Summary

A community forum thread where a user discusses difficulties in forcing ChatGPT to generate text using only a provided list of words due to tokenization.

Key quotes

I’ve since discovered that the way in which ChatGPT tokenizes words and isn’t in fact able to “read” them makes this particularly challenging.

The thread focuses on the limitations of Large Language Models in adhering to strict word-list constraints. The user explores various prompting techniques and Python script generation to overcome tokenization issues.