SST-2 (Stanford Sentiment Treebank, binary):

  • is an NLP single-sentence classification task
  • where the model takes one sentence as input
  • and predicts whether sentiment is positive or negative.

Label meanings

  • positive (1) = the sentence expresses positive sentiment
  • negative (0) = the sentence expresses negative sentiment

Why this task matters

  • tests basic sentiment understanding
  • useful as a clean benchmark for classification fine-tuning

Common metrics

  • accuracy