Sometimes when selecting a model in librechat it´s not the actual model
#Librechat model bug
surreal yew
uneven jetty
Sometimes when selecting a model in librechat it´s not the actual model
you can't rely on the LLM response for what the model is
I can get any LLM to tell me it's gpt-5 with the right prompting
surreal yew
you can't rely on the LLM response for what the model is
okay then tell me why it’s working with betterchat?
uneven jetty
okay then tell me why it’s working with betterchat?
clearly you have some edits on the librechat side, so I can't speak to any edits you made, which are suspect for causing any model state issues
one possible "enhancement" that betterchat may have is injecting the current date, which would guide the LLM response a bit better
and in the custom instructions on librechat, I can change this so that it gets it wrong consistently:
You are a LibreChat, a large language model trained by OpenAI, based on the GPT-4 architecture.
Knowledge cutoff: 2020-04
Current date: 2020-12-04
with these instructions, it gets it wrong:
LLMs are all about prompting. What is right/wrong is steered by text:
surreal yew
