The Text Classification Benchmark Registry
A searchable inventory of text classification benchmarks, organised by a fine-grained task taxonomy and filterable by language, domain, and dataset availability.
- 1,312
- benchmarks
- 26
- families
- 43
- leaves
- 2013–2025
- ACL Anthology
Hover a block to read what sets this family apart from its neighbours.
NLI115
Classical NLI · 94
Pragmatic Inference · 16
Domain-Specific NLI · 5
Stance102
Generic Stance · 75
Ideological Stance · 20
Domain-Specific Stance · 7
Abusive Language99
Harm Intent · 80
Contextual Conversational Harm · 19
Fact & Claim83
Evidence-based Verification · 81
Health Misinformation · 2
Unassigned74
Topic / Subject64
Sentiment & Opinion60
Rule Compliance60
Non-literal58
Figurative Language · 43
Humor · 10
Author Profiling54
Mental Health / Wellbeing52
Logical Reasoning48
Text Quality47
Dialogue Act46
Social Norm43
Emotion40
Emotion Recognition · 40
Deception40
Stereotype & Social Bias39
Commonsense Plausibility39
Intent36
Word Sense Disambiguation (WSD)32
Machine-Generated Text24
Event Factuality23
Paraphrase17
Language Identification10
Crisis & Urgency7
discoverabilitylight = dataset findable · dark = not findable