Text Classification on R8
98.94AccuracyDrop Clause Tsetlin Machine
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Drop Clause Tsetlin MachineDrop Clause Probability=0.752021.05 | 98.94 | — | — | — | — | |
| Lguided-BERT-3layers=32021.04 | 98.28 | — | — | — | — | |
| BERT-large2026.06 | 98.28 | — | — | — | — | |
| RoBerta-GCN2026.04 | 98.2 | — | — | — | — | |
| BERTPre-training=true, Training=true2026.06 | 98.2 | — | — | — | — | |
| G-LossBackbone=RoBERTa-large2026.04 | 98.18 | — | — | — | — | |
| DADGNNSetting=Inductive2021.09 | 98.15 | — | 0.16 | — | — | |
| fastTexth=102026.04 | 98.1 | — | — | — | — | |
| Bert-GCN2026.04 | 98.1 | — | — | — | — | |
| BERT w/ shuf. augm.Setting=Inductive2021.09 | 98.07 | — | 0.21 | — | — | |
| TensorGCNSetting=Transductive2021.09 | 98.04 | — | — | — | — | |
| TensorGCN2026.04 | 98.04 | — | — | — | — | |
| G-LossBackbone=BERT-base2026.04 | 98.04 | — | — | — | — | |
| BERTSetting=Inductive2021.09 | 98.03 | — | 0.24 | — | — | |
| TMPre-training=NTM with BERT-large2026.06 | 97.92 | — | — | — | — | |
| GLoss-O + CEBackbone=BERT-base-uncased2026.04 | 97.91 | — | — | 94.53 | — | |
| DistilBERTSetting=Inductive2021.09 | 97.89 | — | 0.15 | — | — | |
| LLM-Guided TM2026.04 | 97.88 | — | — | — | — | |
| SCL + CEBackbone=BERT-base-uncased2026.04 | 97.88 | — | — | 93.96 | — | |
| TMPre-training=NTM with BERT-base2026.06 | 97.88 | — | — | — | — | |
| BERT-base2026.04 | 97.8 | — | — | — | — | |
| RoBERTa-large2026.04 | 97.8 | — | — | — | — | |
| CEBackbone=RoBERTa-large2026.04 | 97.75 | — | — | 93.46 | — | |
| GLoss-O + CEBackbone=DistilBERT-base2026.04 | 97.71 | — | — | 93.61 | — | |
| TMPre-training=NTM with Top2vec2026.06 | 97.7 | — | — | — | — | |
| GLoss-SQRT + CEBackbone=BERT-base-uncased2026.04 | 97.67 | — | — | 94.34 | — | |
| SCL + CEBackbone=DistilBERT-base2026.04 | 97.65 | — | — | 93.79 | — | |
| CEBackbone=BERT-base-uncased2026.04 | 97.57 | — | — | 93.9 | — | |
| TM with kfeature extension=k nearest words2021.04 | 97.5 | — | — | — | — | |
| TMEmbedding=GloVe2026.04 | 97.5 | — | — | — | — | |
| TMEmbedding=GloVe2026.06 | 97.5 | — | — | — | — | |
| Lguided-BERT-1layers=12021.04 | 97.49 | — | — | — | — | |
| BERT2026.04 | 97.49 | — | — | — | — | |
| G-Loss-O + CEBackbone=RoBERTa-large2026.04 | 97.49 | — | — | 93.04 | — | |
| BERT-base2026.06 | 97.49 | — | — | — | — | |
| CEBackbone=DistilBERT-base2026.04 | 97.48 | — | — | 93.02 | — | |
| GLoss-SQRT + CEBackbone=DistilBERT-base2026.04 | 97.46 | — | — | 93.62 | — | |
| S2GC2021.04 | 97.4 | — | — | — | — | |
| UCPOF2026.02 | 97.4 | — | — | — | — | |
| BERT w/o pos embSetting=Inductive2021.09 | 97.39 | — | 0.2 | — | — | |
| Full RAG2026.02 | 97.35 | — | — | — | — | |
| WideMLPSetting=Inductive2021.09 | 97.27 | — | 0.12 | — | — | |
| G-Loss-SQRT + CEBackbone=RoBERTa-large2026.04 | 97.26 | — | — | 93.02 | — | |
| HeteGCNSetting=Transductive2021.09 | 97.24 | — | 0.51 | — | — | |
| SCL + CEBackbone=RoBERTa-large2026.04 | 97.24 | — | — | 93.11 | — | |
| SGCSetting=Transductive2021.09 | 97.2 | — | 0.1 | — | — | |
| HeteGCNSetting=Inductive2021.09 | 97.17 | — | 0.33 | — | — | |
| TF-IDF + WideMLPSetting=Inductive2021.09 | 97.08 | — | 0.16 | — | — | |
| HyperGATSetting=Inductive2021.09 | 97.07 | — | 0.23 | — | — | |
| TextGCNSetting=Transductive2021.09 | 97.07 | — | — | — | — | |
| TextGCN2026.04 | 97.07 | — | — | — | — | |
| Graph-CNN-C2021.04 | 96.99 | — | — | — | — | |
| TextEnt-fullcomposition=words and entities2018.06 | 96.7 | 91 | — | — | — | |
| WideMLP-2Setting=Inductive2021.09 | 96.61 | — | 1.22 | — | — | |
| TextEnt-wordcomposition=words only2018.06 | 96.5 | 86 | — | — | — | |
| BoE2018.06 | 96.5 | 88.6 | — | — | — | |
| Wikipedia2Vec2018.06 | 96.5 | 88.1 | — | — | — | |
| GloVe + WideMLP-2Setting=Inductive2021.09 | 96.5 | — | 0.14 | — | — | |
| GloVe + WideMLPSetting=Inductive2021.09 | 96.44 | — | 0.08 | — | — | |
| TM with ϕfeature extension=cosine angle theta2021.04 | 96.39 | — | — | — | — | |
| UCPOFModel=Llama-3-8B-Instruct2026.02 | 96.39 | — | — | — | — | |
| Bi-LSTM2021.04 | 96.31 | — | — | — | — | |
| Skip-Gram-Wiki2018.06 | 96.3 | 87.9 | — | — | — | |
| TMrepresentation=BOW2021.04 | 96.16 | — | — | — | — | |
| Tsetlin MachineDrop Clause Probability=02021.05 | 96.16 | — | — | — | — | |
| TM2026.04 | 96.16 | — | — | — | — | |
| TMDescription=Vanilla TM2026.06 | 96.16 | — | — | — | — | |
| fastText2021.04 | 96.13 | — | — | — | — | |
| fastTextSetting=Inductive2021.09 | 96.13 | — | 0.21 | — | — | |
| LSTM (pretrain)pre-training=true2021.04 | 96.09 | — | — | — | — | |
| LSTM (pretrain)Setting=Inductive2021.09 | 96.09 | — | 0.19 | — | — | |
| RNN/LSTMEmbedding=GloVe2026.04 | 96.09 | — | — | — | — | |
| LSTMEmbedding=GloVe2026.06 | 96.09 | — | — | — | — | |
| Full RAGModel=Llama-3-8B-Instruct2026.02 | 96.03 | — | — | — | — | |
| BERT2021.04 | 96.02 | — | — | — | — | |
| CNN-non-staticstatic=false2021.04 | 95.71 | — | — | — | — | |
| TextEnt-entitycomposition=entities only2018.06 | 95.7 | 87.8 | — | — | — | |
| SWEM2021.04 | 95.32 | — | — | — | — | |
| SWEMSetting=Inductive2021.09 | 95.32 | — | 0.26 | — | — | |
| TextCNNPre-training=false, Training=true2026.06 | 95.1 | — | — | — | — | |
| gold shot2026.02 | 94.75 | — | — | — | — | |
| fastText (bigrams)n-gram=bigrams2021.04 | 94.74 | — | — | — | — | |
| fastTexth=102026.06 | 94.74 | — | — | — | — | |
| BOW-SVM2018.06 | 94.7 | 85.1 | — | — | — | |
| CNN-randinitialization=random2021.04 | 94.02 | — | — | — | — | |
| char-CNN2026.04 | 94.02 | — | — | — | — | |
| char-CNN2026.06 | 94.02 | — | — | — | — | |
| TextGCNSetting=Inductive2021.09 | 94 | — | 0.4 | — | — | |
| TF-IDF+LR2021.04 | 93.74 | — | — | — | — | |
| BoW TFIDF2026.04 | 93.74 | — | — | — | — | |
| BoW TFIDF2026.06 | 93.74 | — | — | — | — | |
| LSTMPre-training=false, Training=true2026.06 | 93.7 | — | — | — | — | |
| LSTM2021.04 | 93.68 | — | — | — | — | |
| RNN/LSTM2026.04 | 93.68 | — | — | — | — | |
| LSTM2026.06 | 93.68 | — | — | — | — | |
| gold shotModel=Llama-3-8B-Instruct2026.02 | 93.51 | — | — | — | — | |
| Log. RegressionSetting=Inductive2021.09 | 93.33 | — | — | — | — | |
| LEAM2021.04 | 93.31 | — | — | — | — | |
| baselineModel=Llama-3-8B-Instruct2026.02 | 93.19 | — | — | — | — | |
| W2VPre-training=true, Training=false2026.06 | 93 | — | — | — | — |