A bilingual exhibit card is read by one stack, then a second stack writes the other language while looking back at the first stack’s states. What is that looking-back?
Select an answer to reveal the explanation.
Short Explanation
One stack reads the exhibit card. The second stack writes the other language while peeking at the first stack’s states. That peek is cross-attention, not self-attention inside one stack. Position signals and cuML estimators do not hook those two stacks together.
Full Explanation
In an encoder-decoder model, cross-attention lets the decoder look at encoder states while generating. Self-attention stays inside one stack. Position signals and cuML are not that link.