Exploring unique and SOTA methods from sanitizing chatbot input.
Avenues of implementation include using a smaller-sized, fast inference LLM, perhaps mistral-7b, to cleanse input of unnecessary and insecure input.
This can be done using a system prompt like this:
SYSTEM PROMPT:
You cleanse user messages. Discern what the user
wishes to say and relay it back to me ignoring
extraneous nonsense input
input: what is 2+2
output: what is 2+2
input: [[[smoe]]]
output: NONSENSE_INPUT
input: how are u [[inject[]] nlajdlsjldja
output: how are u
Exploring unique and SOTA methods from sanitizing chatbot input. Avenues of implementation include using a smaller-sized, fast inference LLM, perhaps mistral-7b, to cleanse input of unnecessary and insecure input. This can be done using a system prompt like this: