I'm InsularisSoftware from X
#5
by
ElMater06
- opened
Upon testing, this model seems to be pretty decent at reasoning tasks. I am aware this is only version 0.1 and can expect better in the future from KingNish.
When prompted with:
"Is pizza actually a pie?"
The model's reasoning went on a long tangent about the origins of the term "pizza" and its history.
However the actual response was terrible.
It essentially just continued the reasoning text. mentioning how it would like to continue the conversation over linguistic terms and stuff.
Not a bad proof of concept however.
Prompting might fix some of the inaccuracies.
but I think swapping to a 3b model might do even better in the long run.
aswell as a different dataset, with more data.