Two posters presented at CoSyNe! “Learning a predictive representation of control in RL” and, together with Alexandra Proca and Jan Bauer (equal contribution), “How context representations emerge during training: a linear network perspective”