I will post it in this sub closer to the end of the month:)

Meanwhile be sure to check the Mind’s Eye if you haven’t. Looks like a narrow prototype of abstract thinking for language models.

https://arxiv.org/pdf/2210.05359.pdf