Codesota · Natural Language Processing · Language Modeling · WinograndeTasks/Natural Language Processing/Language Modeling
Language Modeling · benchmark dataset · EN

WinoGrande: An Adversarial Winograd Schema Challenge at Scale.

WinoGrande is a large-scale Winograd-style commonsense reasoning dataset introduced to probe pronoun resolution and robust commonsense understanding. Inspired by the original Winograd Schema Challenge, WinoGrande contains ~44k fill-in-the-blank problems with binary options (right/wrong antecedent). Instances were collected via a careful crowdsourcing pipeline and then filtered with an adversarial filtering algorithm (AFLITE) to reduce dataset-specific statistical biases; roughly half the examples were identified as adversarial in the original release. The benchmark is designed to be harder and less exploitable by spurious correlations than earlier WSC variants; reported human performance is very high (~94%) while state-of-the-art models (as of the paper) were substantially lower. The dataset is available from the authors (allenai) and hosted on the Hugging Face datasets hub.

Paper Submit a result
§ 01 · Leaderboard

Best published scores.

No results indexed yet — be the first to submit a score.

No benchmark results indexed yet
§ 06 · Contribute

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

Submit a result Read submission guide
What a submission needs
  • 01A public checkpoint or API endpoint
  • 02A reproduction script with frozen commit + seed
  • 03Declared evaluation environment (Python, deps)
  • 04One row per metric declared by this dataset
  • 05A contact so we can follow up on discrepancies