Hate Speech Classification on HateXplain
0.763AccuracyCREDENCE-DeBERTa
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CREDENCE-DeBERTaCategory=Encoder, Backbone=DeBERTa, Parameters=66M–184M2026.04 | 0.763 | — | — | — | — | 0.289 | 0.463 | |
| CREDENCE-MistralCategory=LLM, Backbone=Mistral-7B, Parameters=3.8B–8B, Adapter=LoRA2026.04 | 0.752 | — | — | — | — | 0.271 | 0.431 | |
| CREDENCE-Llama-3Category=LLM, Backbone=Llama-3, Parameters=3.8B–8B, Adapter=LoRA2026.04 | 0.749 | — | — | — | — | 0.263 | 0.427 | |
| CREDENCE-RoBERTaCategory=Encoder, Backbone=RoBERTa, Parameters=66M–184M2026.04 | 0.748 | — | — | — | — | 0.267 | 0.445 | |
| Deep EnsembleCategory=General2026.04 | 0.745 | — | — | — | — | 0.195 | 0.267 | |
| CBM + EnsembleCategory=CBM2026.04 | 0.742 | — | — | — | — | 0.184 | 0.271 | |
| Temp. ScalingCategory=General2026.04 | 0.741 | — | — | — | — | 0.112 | 0.198 | |
| MC DropoutCategory=General2026.04 | 0.738 | — | — | — | — | 0.178 | 0.234 | |
| P-CBMCategory=CBM2026.04 | 0.736 | — | — | — | — | 0.171 | 0.258 | |
| CREDENCE-Phi-3Category=LLM, Backbone=Phi-3, Parameters=3.8B–8B, Adapter=LoRA2026.04 | 0.735 | — | — | — | — | 0.245 | 0.398 | |
| Standard CBMCategory=CBM2026.04 | 0.732 | — | — | — | — | 0.145 | 0.221 | |
| CBM + MC DropCategory=CBM2026.04 | 0.73 | — | — | — | — | 0.162 | 0.245 | |
| Evidential DLCategory=General2026.04 | 0.729 | — | — | — | — | 0.134 | 0.198 | |
| Variational Credal2026.02 | 0.724 | 0.06 | 0.33 | 0.68 | 0.74 | — | — | |
| Deep Ensembles2026.02 | 0.721 | 0.8 | 0.26 | 0.34 | 0.67 | — | — | |
| CREDENCE-DistilBERTCategory=Encoder, Backbone=DistilBERT, Parameters=66M–184M2026.04 | 0.721 | — | — | — | — | 0.234 | 0.412 | |
| CBDL2026.02 | 0.72 | 0.59 | 0.3 | 0.4 | 0.7 | — | — | |
| CreINNs2026.02 | 0.719 | 0.62 | 0.29 | 0.38 | 0.69 | — | — | |
| MC Dropout2026.02 | 0.718 | 0.78 | 0.25 | 0.32 | 0.66 | — | — | |
| Semantic Entropy2026.02 | 0.715 | 0.83 | 0.24 | 0.31 | 0.65 | — | — | |
| P(True)2026.02 | 0.712 | 0.75 | 0.23 | 0.3 | 0.64 | — | — |