SQuAD v2.0: leaderboard
Extends SQuAD v1.1 with unanswerable questions that appear plausible, testing whether models can correctly abstain when the passage lacks supporting evidence.
Metric: F1 (%). Source: rajpurkar.github.io. 179 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 89.45 |
Interactive version: theaggregate.ai/benchmark?slug=squad-v2-0 · How It Works · Data refreshed daily, snapshot 2026-10-20.