Rule2: If the task is a judgment that provides a best-so-far solution to a pending goal or question, the achieving level $h$ of the latter is used to increase the priority of the task from $p$ to $or(p, h)$. This reward will facilitate forward inference invoked by backward inference, as it will use the belief associated with the goal or question, without a probabilistic selection of beliefs.
Technical Report
More Info
It seems no such reward is applied