🟢Subheading Benchmarks

Dataset

Name: Subheadings_OneAI_v1.3
Datapoints: 32,561 (tested & reported), 10,000 (published)
Sources: Various sources for news & blogs from various domains (news, politics, tech, finance, etc.)

Results Overview

GrademodelRouge1RougeLRougeLsum
🟢AOurs (One-AI-subheading-v1.7)0.4690.4640.464
🟠FSOTA open-source model 10.1860.1780.178
🟠FSOTA open-source model 20.1330.0440.127
Did we miss a model? suggest one and we'll add it

Note: see our overview on the Rouge metrics

Detailed results

Ours (One-AI-subheadling-v1.7)

SOTA open-source model 1

SOTA open-source model 2


Did this page help you?