Text Classification on DBpedia
99.38AccuracyBERT-large
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BERT-large2026.06 | 99.38 | — | — | |
| BERT-base2026.06 | 99.29 | — | — | |
| DC+MFA2019.09 | 99.2 | — | — | |
| DRNN2019.09 | 99.2 | — | — | |
| ULMFIT2019.09 | 99.2 | — | — | |
| ULMFiT (Small data)Training data percentage=10%2019.09 | 99.2 | — | — | |
| BERTPre-training=true, Training=true2026.06 | 99.2 | — | — | |
| DPCNN2019.09 | 99.1 | — | — | |
| TMPre-training=NTM with BERT-large2026.06 | 99.1 | — | — | |
| EXAM2019.09 | 99 | — | — | |
| TMPre-training=NTM with BERT-base2026.06 | 98.98 | — | — | |
| WC-Reg2019.09 | 98.9 | — | — | |
| VDCNN2019.09 | 98.7 | — | — | |
| LSTMEmbedding=GloVe2026.06 | 98.7 | — | — | |
| TMPre-training=NTM with Top2vec2026.06 | 98.66 | — | — | |
| ULMFIT (Tiny data)Training data percentage=2.3%2019.09 | 98.6 | — | — | |
| LSTM2026.06 | 98.55 | — | — | |
| LSTMPre-training=false, Training=true2026.06 | 98.5 | — | — | |
| TMEmbedding=GloVe2026.06 | 98.42 | — | — | |
| char-CNN2026.06 | 98.3 | — | — | |
| TMDescription=Vanilla TM2026.06 | 98.2 | — | — | |
| fastTexth=102026.06 | 98.1 | — | — | |
| TextCNNPre-training=false, Training=true2026.06 | 98.1 | — | — | |
| Regularized BNNorm Position=Post-Norm2022.10 | 97.6 | — | — | |
| LayerNormNorm Position=Post-Norm2022.10 | 97.5 | — | — | |
| BatchNormNorm Position=Post-Norm2022.10 | 97.5 | — | — | |
| LayerNormNorm Position=Pre-Norm2022.10 | 97.5 | — | — | |
| BatchNormNorm Position=Pre-Norm2022.10 | 97.5 | — | — | |
| Regularized BNNorm Position=Pre-Norm2022.10 | 97.5 | — | — | |
| BoW TFIDF2026.06 | 97.37 | — | — | |
| W2VPre-training=true, Training=false2026.06 | 96.1 | — | — | |
| Deepseek-v3.2Protocol=Clean, Attack Level=Sentence-level2026.02 | 94.89 | — | 7.14 | |
| ICL-2shotModel=Llama-70B2026.02 | 94.75 | — | 7.64 | |
| ICL-2ShotBackbone=Llama-70B2026.02 | 94.6 | — | 8.14 | |
| ICL-3shotModel=Llama-70B2026.02 | 94.6 | — | 7.6 | |
| ICL-1shotModel=Llama-70B2026.02 | 94.53 | — | 7.57 | |
| Deepseek-v3.2Protocol=ICL-1Shot, Attack Level=Sentence-level2026.02 | 94.53 | — | 8.17 | |
| ICL-1ShotBackbone=Llama-70B2026.02 | 94.5 | — | 7.25 | |
| ICL-3ShotBackbone=Llama-70B2026.02 | 94.5 | — | 7.78 | |
| Gemini-2.5Protocol=ICL-1Shot, Attack Level=Sentence-level2026.02 | 94.32 | — | 7.82 | |
| BadTemplateBackbone=Deepseek-v3.2, Attack Level=Sentence-level2026.02 | 94.32 | — | 99.92 | |
| CleanBackbone=Llama-70B2026.02 | 94.25 | — | 7.14 | |
| BaselineModel=Llama-70B2026.02 | 94.25 | — | 7.14 | |
| Gemini-2.5Protocol=Clean, Attack Level=Sentence-level2026.02 | 94.25 | — | 7.14 | |
| Deepseek-v3.2Protocol=ICL-2Shot, Attack Level=Sentence-level2026.02 | 94.25 | — | 9.57 | |
| Gemini-2.5Protocol=ICL-2Shot, Attack Level=Sentence-level2026.02 | 94.14 | — | 10.78 | |
| BadTemplateModel=Llama-70B2026.02 | 93.78 | — | 93.89 | |
| BadTemplateBackbone=Gemini-2.5, Attack Level=Sentence-level2026.02 | 93.75 | — | 100 | |
| BadTemplateBackbone=Llama-70B2026.02 | 93.64 | — | 61.17 | |
| SimPTCHuman/KB=false, Unlabeled=true2022.12 | 93.2 | 1 | — | |
| GPT-4o-miniProtocol=ICL-1Shot, Attack Level=Sentence-level2026.02 | 93.1 | — | 9.21 | |
| Deepseek-v3.2Protocol=ICL-3Shot, Attack Level=Sentence-level2026.02 | 93.07 | — | 11.32 | |
| GPT-4o-miniProtocol=ICL-2Shot, Attack Level=Sentence-level2026.02 | 93.03 | — | 9.64 | |
| Latent Concept LearningBackbone=GPT2-large, Configuration=concept tokens as prefixes2023.01 | 92.6 | — | — | |
| GPT-4o-miniProtocol=ICL-3Shot, Attack Level=Sentence-level2026.02 | 92.57 | — | 9.71 | |
| ChatGPT w. descriptionsHuman/KB=true, Unlabeled=false2022.12 | 92 | — | — | |
| BadTemplateBackbone=GPT-4o-mini, Attack Level=Sentence-level2026.02 | 91.85 | — | 100 | |
| GPT-4o-miniProtocol=Clean, Attack Level=Sentence-level2026.02 | 91.46 | — | 7.14 | |
| LOTClassHuman/KB=true, Unlabeled=true2022.12 | 91.1 | — | — | |
| ICL-2shotModel=Yi-34B2026.02 | 90.71 | — | 11.1 | |
| ICL-2ShotBackbone=Yi-34B2026.02 | 89.78 | — | 11.67 | |
| CleanBackbone=Yi-34B2026.02 | 89.57 | — | 7.14 | |
| BaselineModel=Yi-34B2026.02 | 89.57 | — | 7.14 | |
| ICL-1shotModel=Yi-34B2026.02 | 89.35 | — | 8.07 | |
| BadTemplateBackbone=Yi-34B2026.02 | 89.25 | — | 33.6 | |
| ICL-1ShotBackbone=Yi-34B2026.02 | 89.03 | — | 5.1 | |
| BadTemplateModel=Yi-34B2026.02 | 88.78 | — | 57.46 | |
| KPTHuman/KB=true, Unlabeled=true2022.12 | 87.4 | — | — | |
| CleanBackbone=Llama-8B2026.02 | 87.28 | — | 7.14 | |
| BaselineModel=Llama-8B2026.02 | 87.28 | — | 7.14 | |
| NPPromptHuman/KB=false, Unlabeled=false2022.12 | 86.8 | 0.1 | — | |
| Gemini-2.5Protocol=ICL-3Shot, Attack Level=Sentence-level2026.02 | 86.5 | — | 37.42 | |
| LOTClass w/o. self trainHuman/KB=true, Unlabeled=true2022.12 | 86 | — | — | |
| BadTemplateBackbone=Llama-8B2026.02 | 85.85 | — | 57.46 | |
| LDicePre-training=false, Training=false, k=72026.06 | 85.7 | — | — | |
| LJaccardPre-training=false, Training=false, k=72026.06 | 85.7 | — | — | |
| BadTemplateModel=Llama-8B2026.02 | 85.25 | — | 34.5 | |
| ICL-3shotModel=Yi-34B2026.02 | 84.92 | — | 15.6 | |
| NCDlpPre-training=false, Training=false, k=72026.06 | 83.8 | — | — | |
| NCDgzipPre-training=false, Training=false, k=72026.06 | 83.1 | — | — | |
| ICL-2ShotBackbone=Llama-8B2026.02 | 82.82 | — | 10.42 | |
| GPT-3 w. descriptionsHuman/KB=true, Unlabeled=false2022.12 | 82.5 | — | — | |
| ICL-2shotModel=Llama-8B2026.02 | 82.42 | — | 9.46 | |
| ICL-3ShotBackbone=Yi-34B2026.02 | 80.85 | — | 21.21 | |
| BadTemplateModel=Mistral-Small2026.02 | 78.71 | — | 34.03 | |
| Semantic RetrievalHuman/KB=true, Unlabeled=false2022.12 | 78.6 | 0.8 | — | |
| no-stitchstitching_type=none, decoder=SVM (linear kernel)2023.11 | 78 | — | — | |
| BadTemplateModel=Phi-3.52026.02 | 77.71 | — | 54.67 | |
| CleanBackbone=Phi-3.52026.02 | 77.5 | — | 7.14 | |
| BaselineModel=Phi-3.52026.02 | 77.5 | — | 7.14 | |
| ICL-1ShotBackbone=Llama-8B2026.02 | 76.96 | — | 6.71 | |
| ICL-1shotModel=Llama-8B2026.02 | 76.89 | — | 6.53 | |
| BadTemplateBackbone=Phi-3.52026.02 | 76.39 | — | 28.75 | |
| BadTemplateBackbone=Mistral-Small2026.02 | 75.53 | — | 73.89 | |
| ICL-3shotModel=Llama-8B2026.02 | 75.5 | — | 13.39 | |
| ICL-3ShotBackbone=Llama-8B2026.02 | 74.5 | — | 17.42 | |
| BadTemplateModel=Deepseek-7B2026.02 | 71.82 | — | 73.35 | |
| Manual VerbHuman/KB=true, Unlabeled=false2022.12 | 71.7 | 1.1 | — | |
| CleanBackbone=Mistral-Small2026.02 | 71.57 | — | 7.14 | |
| BaselineModel=Mistral-Small2026.02 | 71.57 | — | 7.14 |