List your company

Natural Language Inference (NLI) (test)

81.17AccuracyNEZHA-wwm (L)
-3.205218.699940.60562.5101Nov 2, 2019Nov 30, 2020Dec 30, 2021Jan 28, 2023Feb 27, 2024Mar 27, 2025Apr 26, 2026

Evaluation results

MethodLinks
NEZHA-wwm (L)Whole Word Masking=true, Backbone Scale=Large2019.1181.17—
ERNIE 2.0 (L)Backbone Scale=Large2019.1181—
NEZHA (L)Backbone Scale=Large2019.1180.44—
ERNIE 2.0 (B)Backbone Scale=Base2019.1179.7—
NEZHA (B)Backbone Scale=Base2019.1179.32—
ZEN (P)Initialization=Pre-trained, Backbone Scale=Base2019.1179.2—
NEZHA-WWM (B)Whole Word Masking=true, Backbone Scale=Base2019.1179.11—
ERNIE 1.0Backbone Scale=Base2019.1178.4—
BERT-WWMWhole Word Masking=true, Backbone Scale=Base2019.1178—
BERT (P)Initialization=Pre-trained, Backbone Scale=Base2019.1177.52—
ZEN (R)Initialization=Random, Backbone Scale=Base2019.1177.03—
SkylineBackbone=multilingual e5, Target Language Adaptation=Korean2026.0475.93—
BERT (R)Initialization=Random, Backbone Scale=Base2019.1175.01—
XITE-reg-regBackbone=multilingual e5, Target Language Adaptation=Korean2026.0457.56—
XITE-reg-ldaBackbone=multilingual e5, Target Language Adaptation=Korean2026.0454.79—
SkylineBackbone=multilingual e5, Target Language Adaptation=Arabic2026.0453.25—
RA-CAShots=250, Selection Strategy=Checkpoint averaging, Run Type=Ensemble2023.0550.50.3
RA-CAShots=50, Selection Strategy=Checkpoint averaging, Run Type=Ensemble2023.05451.6
RA-LASTShots=50, Selection Strategy=Last checkpoint, Run Type=Ensemble2023.0544.62.1
CAShots=50, Selection Strategy=Checkpoint averaging, Run Type=Single Run2023.0544.41.9
XITE-reg-ldaBackbone=multilingual e5, Target Language Adaptation=Arabic2026.0438.84—
TRG-DEVShots=5, Selection Strategy=Early stopping (target dev), Run Type=Single Run2023.0538.31.8
RA-CAShots=5, Selection Strategy=Checkpoint averaging, Run Type=Ensemble2023.0538.31.1
RA-LASTShots=5, Selection Strategy=Last checkpoint, Run Type=Ensemble2023.0538.21
XITE-reg-ldaBackbone=multilingual e5, Target Language Adaptation=Hindi2026.0438.06—
CAShots=5, Selection Strategy=Checkpoint averaging, Run Type=Single Run2023.0537.61.4
GS-LASTShots=5, Selection Strategy=Group selection last, Run Type=Single Run2023.0537.51.8
LASTShots=5, Selection Strategy=Final checkpoint, Run Type=Single Run2023.05371.3
SRC-DEVShots=5, Selection Strategy=Early stopping (source dev), Run Type=Single Run2023.0536.91.3
XITE-reg-regBackbone=multilingual e5, Target Language Adaptation=Arabic2026.0434.35—
XITE-reg-ldaBackbone=multilingual e5, Target Language Adaptation=Urdu2026.0433.81—
XITE-reg-regBackbone=multilingual e5, Target Language Adaptation=Hindi2026.0427.42—
XITE-reg-regBackbone=multilingual e5, Target Language Adaptation=Urdu2026.0423.83—
SkylineBackbone=multilingual e5, Target Language Adaptation=Hindi2026.0418.5—
Baseline-PSBackbone=multilingual e5, Target Language Adaptation=Korean2026.0415.57—
SkylineBackbone=multilingual e5, Target Language Adaptation=Urdu2026.045.63—
Baseline-PSBackbone=multilingual e5, Target Language Adaptation=Urdu2026.040.66—
Baseline-PSBackbone=multilingual e5, Target Language Adaptation=Hindi2026.040.1—
Baseline-PSBackbone=multilingual e5, Target Language Adaptation=Arabic2026.040.04—