Text Classification on DBpedia (test)
0.0061Test Error RateBERT-ITPT-FiT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| BERT-ITPT-FiTBackbone=BERT-Large, Further pre-training=In-Task (ITPT)2019.05 | 0.0061 | — | — | — | — | — | |
| BERT-FiTBackbone=BERT-Large2019.05 | 0.0062 | — | — | — | — | — | |
| BERT-IDPT-FiTconfiguration=BERT + In-Domain Pre-Training + Fine-Tuning2019.05 | 0.0065 | — | — | — | — | — | |
| BERT-CDPT-FiTconfiguration=BERT + Cross-Domain Pre-Training + Fine-Tuning2019.05 | 0.0067 | — | — | — | — | — | |
| BERT-ITPT-FiTconfiguration=BERT + withIn-Task Pre-Training + Fine-Tuning2019.05 | 0.0068 | — | — | — | — | — | |
| BERT-ITPT-FiTBackbone=BERT-Base, Further pre-training=In-Task (ITPT)2019.05 | 0.0068 | — | — | — | — | — | |
| BERT-Featconfiguration=BERT as features2019.05 | 0.007 | — | — | — | — | — | |
| BERT-FiTconfiguration=BERT + Fine-Tuning2019.05 | 0.0071 | — | — | — | — | — | |
| BERT-FiTBackbone=BERT-Base2019.05 | 0.0071 | — | — | — | — | — | |
| ULMFiT2018.01 | 0.008 | — | — | — | — | — | |
| ULMFIT2019.05 | 0.008 | — | — | — | — | — | |
| ULMFiT2019.05 | 0.008 | — | — | — | — | — | |
| CNN2018.01 | 0.0084 | — | — | — | — | — | |
| DPCNN2018.01 | 0.0088 | — | — | — | — | — | |
| DPCNN2019.05 | 0.0088 | — | — | — | — | — | |
| M-ACNNNumber of convolutional filters=Multiple2017.09 | 0.0107 | — | — | — | — | — | |
| Region Emb.2019.05 | 0.011 | — | — | — | — | — | |
| Self-attentive Embedding2017.09 | 0.0114 | — | — | — | — | — | |
| Deep CNNNetwork depth=29 layers2017.09 | 0.0129 | — | — | — | — | — | |
| VDCNN2019.05 | 0.0129 | — | — | — | — | — | |
| D-LSTM2019.05 | 0.013 | — | — | — | — | — | |
| ngrams TFIDFType=TFIDF, N-grams=up to 5-grams2017.09 | 0.0131 | — | — | — | — | — | |
| Deep CNNNetwork depth=9 layers2017.09 | 0.0135 | — | — | — | — | — | |
| ngramsType=bag-of-means, N-grams=up to 5-grams2017.09 | 0.0137 | — | — | — | — | — | |
| Deep CNNNetwork depth=17 layers2017.09 | 0.014 | — | — | — | — | — | |
| Char-level CNN2018.01 | 0.0155 | — | — | — | — | — | |
| Char-level CNN2019.05 | 0.0155 | — | — | — | — | — | |
| M-CNNNumber of convolutional filters=Multiple2017.09 | 0.0166 | — | — | — | — | — | |
| Large word CNNScale=Large, Features per layer=1024, Total layers=62017.09 | 0.0172 | — | — | — | — | — | |
| Small word CNNScale=Small, Features per layer=256, Total layers=62017.09 | 0.0185 | — | — | — | — | — | |
| S-ACNNNumber of convolutional filters=Single2017.09 | 0.0516 | — | — | — | — | — | |
| S-CNNNumber of convolutional filters=Single2017.09 | 0.2235 | — | — | — | — | — | |
| XLNetArchitecture=24-layer, Model size=Large2019.06 | 0.6 | — | — | — | — | — | |
| BERTArchitecture=24-layer, Model size=Large2019.06 | 0.64 | — | — | — | — | — | |
| BERT_LARGESequence Length=5122019.04 | 0.64 | — | — | — | — | — | |
| Mixed VAT2019.06 | 0.7 | — | — | — | — | — | |
| Mixed VATSequence Length=5122019.04 | 0.7 | — | — | — | — | — | |
| ULMFIT2019.06 | 0.8 | — | — | — | — | — | |
| CNN2019.06 | 0.84 | — | — | — | — | — | |
| DPCNN2019.06 | 0.88 | — | — | — | — | — | |
| Black-boxModel backbone=CLIP-base (110M)2026.03 | — | 100 | — | — | — | — | |
| Black-boxModel backbone=CLIP-large (395M)2026.03 | — | 99.3 | — | — | — | — | |
| BM25ChunkContext Window Size (L)=2K, Few-shot Demonstrations=202024.02 | — | 50.16 | — | — | — | — | |
| BM25ChunkContext Window Size (L)=8K, Few-shot Demonstrations=482024.02 | — | 56.57 | — | — | — | — | |
| CT-CBMModel backbone=CLIP-base (110M)2026.03 | — | 99.3 | — | 0.184 | 0.448 | 0.191 | |
| CT-CBMModel backbone=CLIP-large (395M)2026.03 | — | 99 | — | 0.04 | 0.089 | 0.021 | |
| DP SimHash Bucketingprivacy budget (ε)=0.1, number of hyperplanes=24, number of hash tables=42026.05 | — | 64.5 | — | — | — | — | |
| DP SimHash Bucketingprivacy budget (ε)=1, number of hyperplanes=24, number of hash tables=42026.05 | — | 87.9 | — | — | — | — | |
| DP SimHash Bucketingprivacy budget (ε)=3, number of hyperplanes=24, number of hash tables=42026.05 | — | 93.6 | — | — | — | — | |
| DP SimHash Bucketingprivacy budget (ε)=5, number of hyperplanes=24, number of hash tables=42026.05 | — | 95.4 | — | — | — | — | |
| DP SimHash Bucketingprivacy budget (ε)=7, number of hyperplanes=24, number of hash tables=42026.05 | — | 95.9 | — | — | — | — | |
| DP SimHash Bucketingprivacy budget (ε)=8, number of hyperplanes=24, number of hash tables=42026.05 | — | 96.4 | — | — | — | — | |
| Ens. Acc.Model=Claude, Evaluation Protocol=ensemble, epsilon=infinity2023.05 | — | 92.4 | — | — | — | — | |
| f-CBMModel backbone=CLIP-base (110M)2026.03 | — | 99.2 | — | 0.069 | 0.003 | 0.002 | |
| f-CBMModel backbone=CLIP-large (395M)2026.03 | — | 98.9 | — | 0.041 | 0.007 | 0.003 | |
| Indep.-CBMModel backbone=CLIP-base (110M)2026.03 | — | 97.3 | — | 0.045 | 0.027 | 0.004 | |
| Indep.-CBMModel backbone=CLIP-large (395M)2026.03 | — | 98.1 | — | 0.035 | 0.042 | 0.009 | |
| INTRADocContext Window Size (L)=2K, Few-shot Demonstrations=202024.02 | — | 46.82 | — | — | — | — | |
| INTRADocContext Window Size (L)=8K, Few-shot Demonstrations=482024.02 | — | 61.85 | — | — | — | — | |
| KNN-Promptingprivacy budget (ε)=Non-private, number of hyperplanes=24, number of hash tables=42026.05 | — | 98.8 | — | — | — | — | |
| Label-freeModel backbone=CLIP-base (110M)2026.03 | — | 99.2 | — | 1.56 | 0.383 | 0.162 | |
| Label-freeModel backbone=CLIP-large (395M)2026.03 | — | 99.3 | — | 1.773 | 0.317 | 0.116 | |
| Lower BoundModel=Claude, Evaluation Protocol=zero-shot, epsilon=02023.05 | — | 88 | — | — | — | — | |
| MixChunkContext Window Size (L)=2K, Few-shot Demonstrations=202024.02 | — | 40.87 | — | — | — | — | |
| MixChunkContext Window Size (L)=8K, Few-shot Demonstrations=482024.02 | — | 45.94 | — | — | — | — | |
| Non-private One-shot BaselineBackbone=GPT3-Babbage, Protocol=One-shot, Shots=12023.05 | — | 85.6 | — | — | — | — | |
| Private EnsembleBackbone=GPT3-Babbage, Protocol=Ensemble2023.05 | — | 81.6 | — | — | — | — | |
| PromptPATEModel=Claude, epsilon=0.0422023.05 | — | 90.9 | — | — | — | — | |
| PromptPATEBackbone=GPT3-Babbage, Transfer Setting=IID, Public Dataset=dbpedia, Shots=12023.05 | — | 80.3 | 0.194 | — | — | — | |
| PromptPATEBackbone=GPT3-Babbage, Transfer Setting=OOD, Public Dataset=agnews, Shots=12023.05 | — | 74.6 | 0.203 | — | — | — | |
| UNIChunkContext Window Size (L)=2K, Few-shot Demonstrations=202024.02 | — | 36.61 | — | — | — | — | |
| UNIChunkContext Window Size (L)=8K, Few-shot Demonstrations=482024.02 | — | 52.84 | — | — | — | — | |
| Upper BoundModel=Claude, Evaluation Protocol=non-private baseline, epsilon=infinity2023.05 | — | 93.5 | — | — | — | — | |
| Zero-shot BaselineBackbone=GPT3-Babbage, Protocol=Zero-shot, Shots=02023.05 | — | 44.2 | 0 | — | — | — |