List your company
ResearchFormerly WizWand

Sentiment Classification on MR

90.8AccuracyRoBERTa (Gao et al., 2021)
76.614480.297283.9887.6628Aug 25, 2014Jul 13, 2016Jun 2, 2018Apr 21, 2020Mar 11, 2022Jan 29, 2024Dec 18, 2025
Updated 8mo ago

Evaluation results

MethodLinks
RoBERTa (Gao et al., 2021)# Params=1.0x, Evaluation Protocol=Full fine-tuning2022.1290.8
BERT+SCMModel type=Latent semantic tree models, Backbone=BERT2023.0888.16
BERT (2019)Model type=Sequential models2023.0887.65
SBERT2025.1284.9
Kim et al., 2019Model type=Untagged tree (by external parser) models2023.0883.8
NPM# Params=1.0x, Evaluation Protocol=Zero-shot, nonparametric=true2022.1283.7
PICLShot=4-shot, Model=GPT-Neo (2.7B)2023.0583.6
BILSTM+SCMModel type=Latent semantic tree models, Backbone=BiLSTM2023.0883.41
BILSTM (1997)Model type=Sequential models2023.0883.27
CNN+MCFAAdditional Context=Translation, N=ensemble2018.0683.2
AdaSent2018.0683.1
AdaSent2015.0483.1
AdaSentNN category=Other2016.1183.1
TopCNNAdditional Context=Topic, Granularity=ensemble2018.0683
CNN+MCFAAdditional Context=Translation, N=102018.0682.7
CNN+B1Additional Context=Translation, N=ensemble2018.0682.6
CNN+MCFAAdditional Context=Translation, N=12018.0682.3
BRNN2015.0482.3
BLSTM-2DCNNNN category=ours2016.1182.3
DSCNN2018.0682.2
CNN+B2Additional Context=Translation, N=ensemble2018.0682.2
SCD-RoBERTa_baseBackbone=RoBERTa-base2022.0382.17
CNN+B2Additional Context=Translation, N=12018.0682.1
CNN+B2Additional Context=Translation, N=102018.0682.1
Dep-CNN2018.0681.9
CNN+B1Additional Context=Translation, N=12018.0681.9
Liu et al., 2017aModel type=Untagged tree (by external parser) models2023.0881.9
BERT [CLS]-embeddingBackbone=BERT-base, Pooling=CLS2022.0381.83
TopCNNAdditional Context=Topic, Granularity=word2018.0681.7
CNN-Rule-qNetwork role=teacher network, Rule type=but-rule2016.0381.7
Liu et al., 2017bModel type=Untagged tree (by external parser) models2023.0881.7
RoBERTa# Params=1.0x, Evaluation Protocol=Zero-shot2022.1281.7
CNN-Rule-pNetwork role=student network, Rule type=but-rule2016.0381.6
USE_T+CNN (lrn w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=CNN2018.0381.59
CNN-non-staticword vectors=fine-tuned pre-trained (word2vec)2014.0881.5
CNN2018.0681.5
CNN2015.0481.5
DSCNNNN category=Other2016.1181.5
BLSTM-2DPoolingNN category=ours2016.1181.5
USE TUniversal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=None, Transfer Task Model=None2018.0381.44
CNN+B1Additional Context=Translation, N=102018.0681.4
USE_T+DAN (lrn w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=DAN2018.0381.36
USE_T+DAN (w2v w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=DAN2018.0381.32
TopCNNAdditional Context=Topic, Granularity=sentence2018.0681.3
CNNModel variant=non-static2016.0381.3
RoBERTa [CLS]-embeddingBackbone=RoBERTa-base, Pooling=CLS2022.0381.27
USE_T+CNN (w2v w.e.)Universal Encoder=Transformer (USE_T), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=CNN2018.0381.18
CNN-multichannelword vectors=pre-trained (word2vec), channels=22014.0881.1
CNN-multichannel2016.0381.1
IS-BERT_baseBackbone=BERT-base2022.0381.09
CNN-AnaNN category=CNN2016.1181.02
CNN-staticword vectors=static pre-trained (word2vec)2014.0881
BLSTM-AttNN category=ours2016.1181
RoBERTa (Gao et al., 2021)# Params=1.0x, Evaluation Protocol=Zero-shot2022.1280.8
SimCSE-BERT_baseBackbone=BERT-base2022.0380.74
LM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0580.33
STM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0580.09
BLSTMNN category=ours2016.1180
SCLie2025.1279.8
SCNN2025.1279.7
Sent-Parser2014.0879.5
NBSVM2014.0879.4
NB-SVM2015.0479.4
NLM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=LSTM2019.0579.33
LSTM-400Emb.Size=4002019.0579.28
F-Dropout2014.0879.1
CNN-400Emb.Size=4002019.0579.07
LM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0579.06
MV-RNN2014.0879
MNB2014.0879
G-Dropout2014.0879
MNB2015.0479
MV-RecNN2015.0479
G-Dropout2016.0379
MVRNN (2012)Model type=Untagged tree (by external parser) models2023.0879
NLM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0578.89
STM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0578.85
STM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0578.77
TreeLSTM (2015)Model type=Untagged tree (by external parser) models2023.0878.7
Avg. BERT embeddingsBackbone=BERT-base, Pooling=Avg2022.0378.66
LM+TSEDEmb.Size=50, Teacher Model Architecture=LSTM2019.0578.61
NLM+TSED+PT+2LEmb.Size=50, Teacher Model Architecture=CNN2019.0578.6
USE_D+CNN (lrn w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=CNN2018.0378.49
USE_D+CNN (w2v w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence & Word, Embedding Initialization=word2vec skip-gram, Transfer Task Model=CNN2018.0378.2
GPT-2 kNN-LM# Params=2.2x, Evaluation Protocol=Zero-shot2022.1278.2
NLM+TSEDEmb.Size=50, Teacher Model Architecture=CNN2019.0578.1
CNN-50Emb.Size=502019.0578.07
STM+TSEDEmb.Size=50, Teacher Model Architecture=CNN2019.0577.81
CCAE2014.0877.8
RAE2014.0877.7
RAE2015.0477.7
LM+TSEDEmb.Size=50, Teacher Model Architecture=CNN2019.0577.63
USE_D+DAN (lrn w.e.)Universal Encoder=Deep Averaging Network (USE_D), Transfer Level=Sentence, Embedding Initialization=learned on transfer task, Transfer Task Model=DAN2018.0377.57
DPCLie2025.1277.4
Tree-CRF2014.0877.3
GloVe embeddings (avg.)Model=GloVe (avg.)2022.0377.25
CBOW2015.0477.2
RNN2015.0477.2
DPCNN2025.1277.2
LSTM-50Emb.Size=502019.0577.16