Modulating Bottom-Up and Top-Down Visual Processing via Language-Conditional Filters | Read Paper on Bytez