Fake News Detection on LIAR (test)
65.2AccuracyBERT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BERTEvaluation protocol=fine-tuned2026.04 | 65.2 | 72.6 | |
| RoBERTaEvaluation protocol=fine-tuned2026.04 | 64.1 | 72.6 | |
| GPT-4.1-nanoEvaluation protocol=zero-shot2026.04 | 58 | 52.8 | |
| Original PipelineUncertainty Mapping=uncertain → true2026.04 | 52 | 43.4 | |
| Research PipelineUncertainty Mapping=uncertain → true2026.04 | 52 | 44.4 | |
| Research PipelineUncertainty Mapping=uncertain → false2026.04 | 49.5 | 36.3 | |
| Original PipelineUncertainty Mapping=uncertain → false2026.04 | 48.5 | 42.8 | |
| Hybrid CNNMetadata Integrated=All2017.05 | 27.4 | — | |
| CNNsInput=Text only2017.05 | 27 | — | |
| Hybrid CNNMetadata Integrated=Job2017.05 | 25.8 | — | |
| Hybrid CNNMetadata Integrated=State2017.05 | 25.6 | — | |
| SVMsInput=Text only2017.05 | 25.5 | — | |
| Hybrid CNNMetadata Integrated=Speaker2017.05 | 24.8 | — | |
| Hybrid CNNMetadata Integrated=Party2017.05 | 24.8 | — | |
| Logistic RegressionInput=Text only2017.05 | 24.7 | — | |
| Hybrid CNNMetadata Integrated=Context2017.05 | 24.3 | — | |
| Hybrid CNNMetadata Integrated=History2017.05 | 24.1 | — | |
| Hybrid CNNMetadata Integrated=Subject2017.05 | 23.5 | — | |
| Bi-LSTMsInput=Text only2017.05 | 23.3 | — | |
| MajorityInput=Text only2017.05 | 20.8 | — | |
| Original PipelineUncertainty Mapping=none2026.04 | 19 | 19 |