[spacehacker]

Another one:

'Learning to reinforcement learn'

https://arxiv.org/abs/1611.05763