LIAR
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
LIAR-RAW (test)
1.53ChatGPT Meaningfulness Score
7
LIAR RAW
1.85Misleadingness
7
LIAR labeled holdout (test)
64.1AUPC
7
LIAR n=5,000 (5-fold CV)
100AUC
5
LIAR
97.3Politeness
4
LIAR-PLUS
67Accuracy
3
Liar2
0.294AUPRC
3
LIAR (test)
67.73Accuracy
3
LIAR
71ChatGPT Score
2