Concatenating retrieved documents While using the question gets to be infeasible as the sequence duration and sample dimensions increase