Natural Language Inference and Entailment Quiz

Answers and explanations appear as you take the quiz.

1. What are the three standard labels in natural language inference?
   • Entailment, contradiction, neutral
   • Positive, negative, neutral
   • True, false, unknown
   • Subject, verb, object
2. In NLI, what does 'entailment' mean?
   • The two sentences are unrelated
   • The hypothesis must be true given the premise
   • The hypothesis must be false given the premise
   • The sentences are grammatically identical
3. Given premise 'A man is playing guitar on stage,' which hypothesis is a contradiction?
   • A person is performing music
   • The crowd loves the song
   • The stage is empty
   • A musician holds an instrument
4. What is an annotation artifact in an NLI dataset?
   • A model that reasons perfectly with very little data
   • A dataset where every annotator agrees
   • A premise that is always longer than its hypothesis
   • A spurious cue that lets a model guess labels without reasoning about the premise-hypothesis relation
5. Which datasets are widely used to train and evaluate NLI?
   • SNLI and MultiNLI
   • ImageNet and CIFAR
   • MNIST and SVHN
   • WikiText and Penn Treebank
6. How can NLI support a check of a generated claim against source evidence?
   • Measure the generated claim’s length
   • Classify whether the evidence entails, contradicts, or leaves the claim neutral
   • Count unique words only
   • Translate the claim into another language
7. In NLI, what does the neutral label mean?
   • The hypothesis is guaranteed true
   • The hypothesis is guaranteed false
   • The premise neither entails nor contradicts the hypothesis
   • The two sentences are grammatically identical
8. How does a common transformer NLI classifier present a sentence pair to the model?
   • Encode only the hypothesis and discard the premise
   • Convert both sentences into audio first
   • Train a separate model for every word
   • Concatenate the premise and hypothesis with a separator token