AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

WinoGrande: An Adversarial Winograd Schema Challenge at Scale

Keisuke Sakaguchi, Ronan Le Bras, Chandra Bhagavatula, Yejin Choi

AAAI 2020 · Large language models › Retrieval and knowledge · approach: Explanatory and multiple-choice QA benchmarks · 1,090 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Filter predictable instances using AfLite instead of random data reductionWinoGrande · WinoGrande embeddingsKL divergence: 2.53 → 2.51; AfLite → 0.12Improved
Filter twins by PMI differences instead of AfLiteWinoGrande · WinoGrande embeddingsKL divergence: → 2.42Improved
Use AfLite filtering rather than PMI filteringWinoGrande · WinoGrande embeddingsKL divergence: 2.42 → 0.12Improved
Use only local context around the blank instead of the full sentence (ablation)BERT and RoBERTa · WinoGrande debiased testaccuracy (%): 64.9 -> 51.9; 79.1 -> 50.0Hurt
Increase RoBERTa training data from XS to XLRoBERTa · WinoGrande debiased testaccuracy (%): 50.4 -> 79.1Improved
Fine-tune RoBERTa on WinoGrande before WSC evaluation, rather than DPRRoBERTa · WSC testaccuracy (%): 83.1 -> 90.1Improved
Fine-tune RoBERTa on WinoGrande before PDP evaluation, rather than DPRRoBERTa · PDP testaccuracy (%): 86.3 -> 87.5Improved
Sequentially fine-tune on WinoGrande before SuperGLUE-WSCRoBERTa · SuperGLUE-WSC testaccuracy (%): 83.6 -> 85.6Improved
Fine-tune on WinoGrande before DPR evaluationRoBERTa · DPR testaccuracy (%): 91.7 -> 93.1Improved
Fine-tune RoBERTa on WinoGrande before KnowRef evaluation, rather than DPRRoBERTa · KnowRef testaccuracy (%): 84.2 -> 85.6Improved
Sequentially fine-tune on WinoGrande before COPARoBERTa · COPA testaccuracy (%): 86.4 -> 90.6Improved
Fine-tune RoBERTa on WinoGrande rather than DPR for WinogenderRoBERTa · Winogenderabsolute gender accuracy gap: 1.6, 0.9 -> 0.8, 0.8Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy