No one wants to read your imprecise pseudo math in plain text, write in clear sentences
Your premise is that these are not editable models, but you restrict that to a situation that is not true in practice. These models are constantly changing based on human feedback, updated information, config changes, and weight manipulation.
If you follow the conversation around LLMs, one of the top issues people have is inconsistent responses and the models changing under their prompts and breaking them.
Your premise is that these are not editable models, but you restrict that to a situation that is not true in practice. These models are constantly changing based on human feedback, updated information, config changes, and weight manipulation.
If you follow the conversation around LLMs, one of the top issues people have is inconsistent responses and the models changing under their prompts and breaking them.