Sentiment Analysis on SST-5 (test)
62.27AccuracyCAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| CAPOModel=Llama-3.3-70B2025.04 | 62.27 | |
| OPROModel=Llama-3.3-70B2025.04 | 60.87 | |
| EvoPromptGAModel=Llama-3.3-70B2025.04 | 60.53 | |
| CAPOModel=Mistral-Small-24B2025.04 | 60.2 | |
| Fine-tuning*Backbone=RoBERTa-large, Setting=Supervised2023.05 | 59.9 | |
| CAPOModel=Qwen2.5-32B2025.04 | 59.07 | |
| EvoPromptGAModel=Qwen2.5-32B2025.04 | 58.6 | |
| InitialModel=Llama-3.3-70B2025.04 | 58.47 | |
| OPROModel=Qwen2.5-32B2025.04 | 57 | |
| InitialModel=Qwen2.5-32B2025.04 | 56.68 | |
| AdamWBackbone=RoBERTa-large, Optimization Method=AdamW, Finetuning Protocol=Few-shot, Iterations=10K2025.11 | 56.6 | |
| CAPOFramework=promptolution, LLM=Gemma-3-27B-it2025.12 | 56.3 | |
| OPROFramework=promptolution, LLM=Gemma-3-27B-it2025.12 | 56 | |
| AutoDiffFramework=AdalFlow, LLM=Gemma-3-27B-it2025.12 | 55.7 | |
| LoRA-nBARBackbone=RoBERTa-large, Number of Parameters=355M, Evaluation Protocol=few-shot2024.10 | 55 | |
| EvoPromptGAModel=Mistral-Small-24B2025.04 | 54.93 | |
| BCN+ELMo2018.02 | 54.7 | |
| ELMo + BaselineModel Configuration=ELMo enhanced2018.02 | 54.7 | |
| LoRA-oBARBackbone=RoBERTa-large, Number of Parameters=355M, Evaluation Protocol=few-shot2024.10 | 54.5 | |
| LoRA-SAMBackbone=RoBERTa-large, Number of Parameters=355M, Evaluation Protocol=few-shot2024.10 | 54.2 | |
| BCN+Char+CoVeembeddings=GloVe, character n-gram, CoVe2017.08 | 53.7 | |
| BCN+Char+CoVe2018.02 | 53.7 | |
| McCann et al. (2017)Model Configuration=Previous SOTA2018.02 | 53.7 | |
| EvoPromptFramework=promptolution, LLM=Gemma-3-27B-it2025.12 | 53.3 | |
| OPROModel=Mistral-Small-24B2025.04 | 53.2 | |
| NTI2017.08 | 53.1 | |
| NTI2018.02 | 53.1 | |
| TE-LSTM2017.08 | 52.6 | |
| LSTM-CNN2017.08 | 52.4 | |
| LSTM-CNN2018.02 | 52.4 | |
| LoRABackbone=RoBERTa-large, Number of Parameters=355M, Evaluation Protocol=few-shot2024.10 | 52.3 | |
| DMN2017.08 | 52.1 | |
| DMN2015.06 | 52.1 | |
| DMN2018.02 | 52.1 | |
| MeZO+MomentumBackbone=RoBERTa-large, Optimization Method=MeZO+Momentum (Mom.), Finetuning Protocol=Few-shot, Iterations=10K, Smoothing parameter=10^-32025.11 | 51.9 | |
| MVN2017.08 | 51.5 | |
| d-TBCNNGroup=TBCNNs2015.04 | 51.4 | |
| Our baselineModel Configuration=Baseline2018.02 | 51.4 | |
| Molding-CNN2015.11 | 51.2 | |
| T-CNN2016.01 | 51.2 | |
| Constituency Tree-LSTMword embeddings=Glove vectors, tuned2015.02 | 51 | |
| Constituency Tree-LSTM2015.11 | 51 | |
| CT-LSTM2015.06 | 51 | |
| CT-LSTMmemory size=1682016.01 | 51 | |
| Tree LSTM (variant 2)Group=RNNs2015.04 | 50.6 | |
| CNN with self-attentionconnection=Dense2018.08 | 50.6 | |
| Dynamic Self-Attentionmode=Single DSA2018.08 | 50.6 | |
| c-TBCNNGroup=TBCNNs2015.04 | 50.4 | |
| BiLSTM with self-attention2018.08 | 50.4 | |
| Tree LSTM (variant 3)Group=RNNs2015.04 | 49.9 | |
| DRNN2015.02 | 49.8 | |
| Deep RNNGroup=RNNs2015.04 | 49.8 | |
| DRNN2015.11 | 49.8 | |
| DRNN2015.06 | 49.8 | |
| DRNN2016.01 | 49.8 | |
| Constituency Tree-LSTMword embeddings=Glove vectors, fixed2015.02 | 49.7 | |
| BiLSTM2018.08 | 49.5 | |
| MeZOBackbone=RoBERTa-large, Optimization Method=MeZO, Finetuning Protocol=Few-shot, Iterations=10K, Smoothing parameter=10^-32025.11 | 49.3 | |
| C-LSTM2015.11 | 49.2 | |
| Bidirectional LSTM2015.02 | 49.1 | |
| bi-LSTMGroup=Recurrent2015.04 | 49.1 | |
| ConMeZOBackbone=RoBERTa-large, Optimization Method=ConMeZO, Finetuning Protocol=Few-shot, Iterations=10K, Smoothing parameter=10^-32025.11 | 48.9 | |
| TE-MNLIBackbone=RoBERTa-large, Labeled=true, Setting=Zero-shot2023.05 | 48.8 | |
| Paragraph-Vec2015.02 | 48.7 | |
| Paragraph vectorGroup=Vector2015.04 | 48.7 | |
| Paragraph Vector2015.11 | 48.7 | |
| PVec2015.06 | 48.7 | |
| PV2016.01 | 48.7 | |
| InitialModel=Mistral-Small-24B2025.04 | 48.69 | |
| DCNN2015.02 | 48.5 | |
| 2-layer Bidirectional LSTM2015.02 | 48.5 | |
| Deep CNNGroup=CNNs2015.04 | 48.5 | |
| DCNN2015.11 | 48.5 | |
| DCNN2015.06 | 48.5 | |
| DCNN2016.01 | 48.5 | |
| Dependency Tree-LSTMword embeddings=Glove vectors, tuned2015.02 | 48.4 | |
| Dependency Tree-LSTM2015.11 | 48.4 | |
| CNN-non-static2015.02 | 48 | |
| Non-staticGroup=CNNs2015.04 | 48 | |
| Tree LSTM (variant 1)Group=RNNs2015.04 | 48 | |
| CNN-non-static2015.11 | 48 | |
| CNN-MC2016.01 | 48 | |
| CNNtype=non-static2018.08 | 48 | |
| 2-layer LSTMNlayers=2, memory size=1682016.01 | 47.9 | |
| Bi-LSTMimplementation=author implementation2015.11 | 47.8 | |
| TE-MNLIBackbone=BART-large, Labeled=true, Setting=Zero-shot2023.05 | 47.7 | |
| LSTMNmemory size=1682016.01 | 47.6 | |
| CNN-multichannel2015.02 | 47.4 | |
| MultichannelGroup=CNNs2015.04 | 47.4 | |
| CNN-multichannel2015.11 | 47.4 | |
| CNN-MC2015.06 | 47.4 | |
| SSTuning-largeBackbone=RoBERTa-large, Labeled=false, Setting=Zero-shot2023.05 | 46.9 | |
| LSTMimplementation=author implementation2015.11 | 46.6 | |
| LSTM2015.02 | 46.4 | |
| LSTMmemory size=1682016.01 | 46.4 | |
| 2-layer LSTM2015.02 | 46 | |
| 2-layer LSTMlayers=2, memory size=1682016.01 | 46 | |
| LSTMGroup=RNNs2015.04 | 45.8 | |
| RNTN2015.02 | 45.7 | |
| TensorGroup=RNNs2015.04 | 45.7 |