Which of the following statements accurately describe the capabilities and functions of sequence transduction models? Select all that apply.

Sequence transduction models take a sequence of inputs and produce a new sequence as output, handling cases where the input and output lengths differ. They learn how to align elements of the input to the output, so they can generate words in translation or transcribe speech. The models can be trained on pairs of sequences, learning to predict the next element while conditioning on all previous predictions. For example, a neural machine translation system reads a sentence in English and outputs a sentence in French, learning to map the input sequence to a new output sequence. This ability makes them useful for tasks such as speech recognition, text summarization, and machine translation.