WinoGrande generalizes the classic Winograd Schema Challenge by providing many more examples that are adversarially filtered. Each item requires choosing the correct referent for an ambiguous pronoun based on commonsense and world knowledge.
Because distractors are carefully constructed, WinoGrande discourages heuristics and highlights whether models genuinely track entities and causal relations in text.
Have a question? Noticed something wrong? Let us know.
A large-scale pronoun resolution and coreference benchmark designed to reduce annotation artifacts and emphasize commonsense.