Aspect-Based Sentiment Analysis on SemEval Task 4 Subtask 2 Restaurant domain 2014 (test)
91.76AccuracyArctic-ABSA Decoder
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Arctic-ABSA DecoderNumber of Parameters=8B, Model Architecture=Decoder2026.01 | 91.76 | — | — | |
| Flan-T5+RVISA_gNumber of Parameters=11B2026.01 | 91.52 | — | — | |
| LSA_E-X-DeBERTaNumber of Parameters=418M2026.01 | 90.98 | — | — | |
| Arctic-ABSA Encoder-thinking +synthNumber of Parameters=395M, Model Architecture=Encoder, Thinking Mode=true, Data Augmentation=SemEval14-synth2026.01 | 90.05 | — | — | |
| DPLMethod Category=Ours, Backbone=BERT2022.03 | 89.54 | 84.86 | — | |
| Arctic-ABSA Encoder +synthNumber of Parameters=395M, Model Architecture=Encoder, Data Augmentation=SemEval14-synth2026.01 | 89.34 | — | — | |
| Flan-T5+RVISA_gNumber of Parameters=250M2026.01 | 89.29 | — | — | |
| Arctic-ABSA Encoder-thinkingNumber of Parameters=395M, Model Architecture=Encoder, Thinking Mode=true, Data Augmentation=None2026.01 | 88.71 | — | — | |
| Arctic-ABSA EncoderNumber of Parameters=395M, Model Architecture=Encoder, Data Augmentation=None2026.01 | 88.35 | — | — | |
| Instruct-ABSA-2Number of Parameters=200M2026.01 | 88.03 | — | — | |
| Rietzler et al. (2019)Method Category=BERT, Backbone=BERT2022.03 | 87.89 | 81.05 | — | |
| CMV-Fuse2025.12 | 87.76 | — | 81.99 | |
| Pang et al. (2021)Method Category=BERT, Backbone=BERT2022.03 | 87.66 | 82.97 | — | |
| BiSyn-GAT2025.12 | 87.49 | — | 81.63 | |
| SSEGCN2025.12 | 87.31 | — | 81.09 | |
| Li et al. (2021)Method Category=BERT, Backbone=BERT2022.03 | 87.13 | 81.16 | — | |
| Dual GCN2025.12 | 87.13 | — | 81.16 | |
| KE-IGCN2025.12 | 86.7 | — | 81.05 | |
| R-GAT2025.12 | 86.6 | — | 81.35 | |
| DGEDTCategory=Labeled Data Dependent Methods2023.10 | 86.3 | — | 80 | |
| dotGCNCategory=Labeled Data Dependent Methods2023.10 | 86.2 | — | 80.5 | |
| dotGCN2025.12 | 86.16 | — | 80.49 | |
| Bai et al. (2020)Method Category=BERT, Backbone=BERT, replication=true2022.03 | 86.04 | 80.27 | — | |
| BERT2025.12 | 85.62 | — | 78.28 | |
| Liang et al. (2020)Method Category=Auxiliary2022.03 | 84.93 | 76.66 | — | |
| He et al. (2019b)Method Category=Auxiliary2022.03 | 83.89 | 75.66 | — | |
| Chen and Qian (2019)Method Category=Auxiliary2022.03 | 79.55 | 71.41 | — | |
| He et al. (2018)Method Category=Auxiliary2022.03 | 78.73 | 68.63 | — | |
| COLACategory=Zero-shot Methods2023.10 | 74.1 | — | 65.7 | |
| GPT-3.5 TurboCategory=Zero-shot Methods2023.10 | 70.6 | — | 59.7 |