What is steady state learning in the context of boundedly rational expectation modeling and how does it differ with other theories of learning?

What is steady state learning in the context of boundedly rational expectation modeling? E.g. see Evans et al. (2009)

They define it as: "Under steady-state learning, agents use past data to estimate the mean interest rate, which is then used as the forecast of future interest rates for all horizons".

However, also under adaptive learning past data are used to produce forecasts. Is SS learning just a synonym for infinite horizon adaptive learning? Or does it refer to a form of learning where rational expectations are replaced by boundely rational expectations in the steady state linear approximated solutions of the model? Rather than replacing them in the constituting equations directly and see explore determinacy/stability of SS solutions as such?

I would like to know what the main differences are with adaptive learning, Euler equation learning, infinite horizon adaptive learning, etc. Adaptive learning could also be a trend following rule, an anchor and adjustment rule, etc. So I wondering what has been used here. In many policy oriented papers the specific adaptive rule that is employed by the agents is hard to find in the paper actually.

Evans, G.W. , Honkapohja, S. , Mitra, K. , 2009. Anticipated fiscal policy and adaptive learning. J. Monet. Econ. 56 (7), 930–953 .