Ideally yes but that would not work with the model latency. But it is already partly covered when a ghost enters pacmans corridor he is asked again. Here's how the decisions work in detail https://github.com/opper-ai/jevman-benchmark/blob/main/READM...