Teacher Forcing in Sequence Models Quiz

Answers and explanations appear as you take the quiz.

1. During teacher forcing, what is fed as the input at each decoding step?
   • The ground-truth previous token
   • The model's own previous prediction
   • A random token
   • The final output token
2. During teacher-forced training, what target is paired with decoder input y_(t−1)?
   • The previous ground-truth token
   • The next target token y_t
   • A token sampled from the validation set
   • The maximum-probability output from an earlier training run
3. The train-test mismatch caused by teacher forcing is known as:
   • Vanishing gradients
   • Mode collapse
   • Exposure bias
   • Label leakage
4. In a Transformer decoder, what mechanism lets teacher-forced training run in parallel across positions?
   • Dropout
   • Batch normalization
   • Beam search
   • A causal (look-ahead) attention mask
5. At inference time, why can't teacher forcing be used?
   • There are no ground-truth future tokens to feed
   • The loss function is unavailable
   • The GPU is too slow
   • Attention masks are disabled
6. Which loss is typically used at each step during teacher-forced training?
   • Mean squared error on raw logits
   • Cross-entropy between the predicted distribution and the true next token
   • Hinge loss on embeddings
   • Cosine similarity loss
7. The target sequence is “red, blue, green.” Under teacher forcing, which token is supplied when training the prediction of “green”?
   • The model’s prediction for “green”
   • The future token after “green”
   • The ground-truth previous token “blue”
   • A randomly sampled vocabulary item
8. At decoder position t, which tokens must a causal attention mask block?
   • The source sequence
   • All earlier target tokens
   • The current position’s own input
   • Target tokens at positions later than t