Replay-Buffer Transfer

For 12-H2O under combined depolarizing noise, transfer achieved the highest composite score and large reductions in steps and gate counts. Buffer-only noiseless-to-noisy transfer produced consistent gains in the reported experiments. Transfer experiments used OptCRLQAS with uniform replay to isolate the effect of transfer. The buffer-transfer method copies source-buffer transitions into the target buffer without relabeling, filtering, or network-weight transfer. Replay-buffer transfer relies on noiseless and noisy tasks sharing state and action spaces. Replay-buffer transfer may fail when source and target tasks differ substantially in topology, gate sets, or encodings.