Text Classification on 20News
96.77AccuracyMTL
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MTLScenario=Non-continual Learning (SDL), Backbone=BERT2021.12 | 96.77 | 96.77 | — | — | — | — | |
| BLSTM-2DCNNNN category=ours2016.11 | 96.5 | — | — | — | — | — | |
| RCNNNN category=RNN2016.11 | 96.49 | — | — | — | — | — | |
| SDLScenario=Non-continual Learning (SDL), Backbone=BERT2021.12 | 96.49 | 96.48 | — | — | — | — | |
| SDLScenario=Non-continual Learning (SDL), Backbone=BERT (Frozen)2021.12 | 96.49 | 96.48 | — | — | — | — | |
| SDLScenario=Non-continual Learning (SDL), Backbone=Adapter-BERT2021.12 | 96.2 | 96.19 | — | — | — | — | |
| CTR (forward)Scenario=Continual Learning (CL), Backbone=BERT2021.12 | 95.63 | 95.62 | — | — | — | — | |
| BLSTM-2DPoolingNN category=ours2016.11 | 95.5 | — | — | — | — | — | |
| CTRScenario=Continual Learning (CL), Backbone=BERT2021.12 | 95.25 | 95.23 | — | — | — | — | |
| HATScenario=Continual Learning (CL), Backbone=Adapter-BERT2021.12 | 95.22 | 95.21 | — | — | — | — | |
| CATScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 95.17 | 95.16 | — | — | — | — | |
| B-CLScenario=Continual Learning (CL), Backbone=BERT2021.12 | 95.07 | 95.04 | — | — | — | — | |
| SDLScenario=Non-continual Learning (SDL), Backbone=W2V2021.12 | 94.72 | 94.72 | — | — | — | — | |
| UCLScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 94.65 | 94.63 | — | — | — | — | |
| BLSTM-AttNN category=ours2016.11 | 94.6 | — | — | — | — | — | |
| BLSTMNN category=ours2016.11 | 94 | — | — | — | — | — | |
| HATScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 93.51 | 92.93 | — | — | — | — | |
| A-GEMScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 93.31 | 92.95 | — | — | — | — | |
| EWCScenario=Continual Learning (CL), Backbone=W2V2021.12 | 91.86 | 91.8 | — | — | — | — | |
| NFHScenario=Continual Learning (CL), Backbone=W2V2021.12 | 90.74 | 90.59 | — | — | — | — | |
| A-GEMScenario=Continual Learning (CL), Backbone=W2V2021.12 | 90.72 | 90.6 | — | — | — | — | |
| UCLScenario=Continual Learning (CL), Backbone=W2V2021.12 | 90.61 | 90.46 | — | — | — | — | |
| DER++Scenario=Continual Learning (CL), Backbone=W2V2021.12 | 89.28 | 89.19 | — | — | — | — | |
| SGCSetting=Transductive2021.09 | 88.5 | — | — | 0.1 | — | — | |
| RMDLNumber of RDLs=152018.05 | 87.91 | — | — | — | — | — | |
| TensorGCNSetting=Transductive2021.09 | 87.74 | — | — | — | — | — | |
| RMDLNumber of RDLs=92018.05 | 87.62 | — | — | — | — | — | |
| TR-BERTBackbone=BERT-base, Padding=No2022.05 | 87.4 | — | — | — | 4.22 | — | |
| BERTSetting=Inductive2021.09 | 87.21 | — | — | 0.18 | — | — | |
| HeteGCNSetting=Transductive2021.09 | 87.15 | — | — | 0.15 | — | — | |
| RMDLNumber of RDLs=32018.05 | 86.73 | — | — | — | — | — | |
| BERT_baseBackbone=BERT-base, Padding=No2022.05 | 86.7 | — | — | — | 1 | — | |
| HyperGATSetting=Inductive2021.09 | 86.62 | — | — | 0.16 | — | — | |
| DNN2018.05 | 86.5 | — | — | — | — | — | |
| POWER-BERTBackbone=BERT-base, Padding=Sequence2022.05 | 86.5 | — | — | — | 2.91 | — | |
| BERT w/ shuf. augm.Setting=Inductive2021.09 | 86.46 | — | — | 0.42 | — | — | |
| TextGCNSetting=Transductive2021.09 | 86.34 | — | — | — | — | — | |
| DistilBERTSetting=Inductive2021.09 | 86.24 | — | — | 0.26 | — | — | |
| TranskimmerBackbone=BERT-base, Padding=No2022.05 | 86.1 | — | — | — | 5.27 | — | |
| SVM(TF-IDF)Representation=TF-IDF2018.05 | 86 | — | — | — | — | — | |
| LDS-PNWays=5, Shots=5-shot2026.02 | 85.4 | — | — | — | — | — | |
| SWEMSetting=Inductive2021.09 | 85.16 | — | — | 0.29 | — | — | |
| TARTWays=5, Shots=5-shot2026.02 | 84.9 | — | — | — | — | — | |
| HeteGCNSetting=Inductive2021.09 | 84.59 | — | — | 0.14 | — | — | |
| SVM2018.05 | 84.57 | — | — | — | — | — | |
| OWMScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 84.54 | 82.73 | — | — | — | — | |
| TextEnt-fullcomposition=words and entities2018.06 | 84.5 | — | 83.9 | — | — | — | |
| TF-IDF + WideMLPSetting=Inductive2021.09 | 84.2 | — | — | 0.16 | — | — | |
| RNN2018.05 | 83.75 | — | — | — | — | — | |
| Log. RegressionSetting=Inductive2021.09 | 83.7 | — | — | — | — | — | |
| TextEnt-wordcomposition=words only2018.06 | 83.6 | — | 82.8 | — | — | — | |
| WideMLPSetting=Inductive2021.09 | 83.31 | — | — | 0.22 | — | — | |
| NFHScenario=Continual Learning (CL), Backbone=BERT (Frozen)2021.12 | 83.28 | 81.81 | — | — | — | — | |
| Shot-DEWays=5, Shots=5-shot2026.02 | 83.2 | — | — | — | — | — | |
| Way-DEWays=5, Shots=5-shot2026.02 | 83.2 | — | — | — | — | — | |
| TextEnt-entitycomposition=entities only2018.06 | 83.1 | — | 82.4 | — | — | — | |
| BoE2018.06 | 83.1 | — | 82.7 | — | — | — | |
| ProtoVerbWays=5, Shots=5-shot2026.02 | 83.1 | — | — | — | — | — | |
| CNN2018.05 | 82.91 | — | — | — | — | — | |
| Wikipedia2Vec2018.06 | 82.9 | — | 82.3 | — | — | — | |
| Skip-Gram-Wiki2018.06 | 82.2 | — | 81.5 | — | — | — | |
| SPCNetWays=5, Shots=5-shot2026.02 | 81.7 | — | — | — | — | — | |
| Naïve Bayes Classifier2018.05 | 81.67 | — | — | — | — | — | |
| BERT w/o pos embSetting=Inductive2021.09 | 81.47 | — | — | 0.49 | — | — | |
| WideMLP-2Setting=Inductive2021.09 | 81.02 | — | — | 0.23 | — | — | |
| TextGCNSetting=Inductive2021.09 | 80.88 | — | — | 0.54 | — | — | |
| ContrastNetWays=5, Shots=5-shot2026.02 | 80.5 | — | — | — | — | — | |
| EWCScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 80.26 | 78.6 | — | — | — | — | |
| SRKScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 79.64 | 77.89 | — | — | — | — | |
| fastTextSetting=Inductive2021.09 | 79.38 | — | — | 0.3 | — | — | |
| BOW-SVM2018.06 | 79 | — | 78.3 | — | — | — | |
| MLADAWays=5, Shots=5-shot2026.02 | 77.8 | — | — | — | — | — | |
| LaSAMLWays=5, Shots=5-shot2026.02 | 77.7 | — | — | — | — | — | |
| LoRATrainable Params=30,670,848, Backbone Model=Qwen3 8B2026.04 | 77.03 | — | — | — | — | — | |
| ShadowPEFTTrainable Params=29,118,720, Backbone Model=Qwen3 8B, Inference Mode=Shadow-attached2026.04 | 76.99 | — | — | — | — | — | |
| GloVe + WideMLPSetting=Inductive2021.09 | 76.8 | — | — | 0.11 | — | — | |
| DoRATrainable Params=31,039,488, Backbone Model=Qwen3 8B2026.04 | 76.78 | — | — | — | — | — | |
| ReLoRAUpdate Source=OpenOrca, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.6 | — | — | — | — | — | |
| PortLLM+FTUpdate Source=OpenOrca, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.4 | — | — | — | — | — | |
| ReLoRAUpdate Source=OpenPlatypus, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.4 | — | — | — | — | — | |
| GloVe + WideMLP-2Setting=Inductive2021.09 | 76.33 | — | — | 0.18 | — | — | |
| ShadowPEFT w/ pretrained shadowTrainable Params=454,991,872, Backbone Model=Qwen3 8B + 0.5B Shadow, Inference Mode=Shadow-attached, Shadow Model Initialization=pretrained2026.04 | 76.31 | — | — | — | — | — | |
| ReLoRAUpdate Source=AlpacaGPT4, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.3 | — | — | — | — | — | |
| LoRA-ScratchUpdate Source=OpenOrca, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.2 | — | — | — | — | — | |
| ORALUpdate Source=OpenOrca, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.1 | — | — | — | — | — | |
| PortLLM+FTUpdate Source=AlpacaGPT4, Evaluation Protocol=Task-adaptation Methods2026.05 | 76.1 | — | — | — | — | — | |
| ShadowPEFT w/ random shadowTrainable Params=454,991,872, Backbone Model=Qwen3 8B + 0.5B Shadow, Inference Mode=Shadow-attached, Shadow Model Initialization=random2026.04 | 75.88 | — | — | — | — | — | |
| PortLLM+FTUpdate Source=OpenPlatypus, Evaluation Protocol=Task-adaptation Methods2026.05 | 75.8 | — | — | — | — | — | |
| ShadowPEFTTrainable Params=23,442,176, Backbone Model=Qwen3 4B, Inference Mode=Shadow-attached2026.04 | 75.73 | — | — | — | — | — | |
| LoRA-ScratchUpdate Source=AlpacaGPT4, Evaluation Protocol=Task-adaptation Methods2026.05 | 75.6 | — | — | — | — | — | |
| ORALUpdate Source=AlpacaGPT4, Evaluation Protocol=Task-adaptation Methods2026.05 | 75.5 | — | — | — | — | — | |
| LSTM (pretrain)Setting=Inductive2021.09 | 75.43 | — | — | 1.72 | — | — | |
| DoRATrainable Params=23,906,304, Backbone Model=Qwen3 4B2026.04 | 75.31 | — | — | — | — | — | |
| LDS-PNWays=5, Shots=1-shot2026.02 | 75.3 | — | — | — | — | — | |
| LoRATrainable Params=23,592,960, Backbone Model=Qwen3 4B2026.04 | 75.23 | — | — | — | — | — | |
| ORALUpdate Source=OpenPlatypus, Evaluation Protocol=Task-adaptation Methods2026.05 | 74.6 | — | — | — | — | — | |
| LoRA-ScratchUpdate Source=OpenPlatypus, Evaluation Protocol=Task-adaptation Methods2026.05 | 74.3 | — | — | — | — | — | |
| TARTWays=5, Shots=1-shot2026.02 | 73.2 | — | — | — | — | — | |
| KANScenario=Continual Learning (CL), Backbone=BERT (frozen)2021.12 | 73.07 | 69.97 | — | — | — | — | |
| Detached Shadow Only (pretrained)Trainable Params=454,991,872, Backbone Model=Qwen3 8B + 0.5B Shadow, Inference Mode=Detached, Shadow Model Initialization=pretrained2026.04 | 72.63 | — | — | — | — | — |