As it stands, agents continue running while their attempts are still pending. As a result, an agent may submit multiple attempts before receiving any scores, which can cause it to keep exploring paths that do not improve performance.
As a tradeoff between cost / time we can stop an agent that submitted an attempt until the score is returned.
As it stands, agents continue running while their attempts are still pending. As a result, an agent may submit multiple attempts before receiving any scores, which can cause it to keep exploring paths that do not improve performance.
As a tradeoff between cost / time we can stop an agent that submitted an attempt until the score is returned.