SQuAD 1.1 (F1): leaderboard
SQuAD 1.1 (Rajpurkar et al., 2016): extractive reading comprehension on Wikipedia passages, where every question is answerable by a span of the passage. This board is F1, which gives partial credit for word overlap with a reference answer, from the archived Papers With Code table of reported scores; higher is better.
Metric: F1. Source: huggingface.co. Saturation forecast: Rough model projection: around 2026. 198 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | {ANNA} (single model) | 95.72 | |
| 2 | LUKE 483M | 95.4 | |
| 3 | LUKE (single model) | 95.38 | |
| 4 | XLNet (single model) | 95.08 | |
| 5 | XLNET-123 (single model) | 94.93 | |
| 6 | XLNET-123++ (single model) | 94.9 | |
| 7 | XLNET-123+ (single model) | 94.86 | |
| 8 | SpanBERT (single model) | 94.64 | |
| 9 | Unnamed submission by NMC | 94.58 | |
| 10 | BERTSP (single model) | 94.58 |
Interactive version: theaggregate.ai/benchmark?slug=squad-1-1-f1 · How It Works · Data refreshed daily, snapshot 2026-11-05.