Aspect-based Sentiment Classification on Rest SemEval-2016 (test)
94.62AccuracyABSA-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ABSA-R12026.04 | 94.62 | 81.69 | |
| w/o SARAblation=without sentiment-aligned reward2026.04 | 93.69 | 73.58 | |
| InstructSFT2026.04 | 93.38 | 72.7 | |
| w/o PRSAblation=without performance-driven rejection sampling2026.04 | 93.38 | 73.18 | |
| CKG-BERTBackbone=BERT2026.04 | 92.87 | 81.95 | |
| KGAN2026.04 | 92.34 | 81.31 | |
| ASHGAT-BERTBackbone=BERT2026.04 | 90.75 | 77.57 | |
| MGGCN-BERTBackbone=BERT2026.04 | 89.66 | 73.99 | |
| MCRF-SA2020.10 | 89.51 | 75.92 | |
| InstructABSA22026.04 | 89.43 | — | |
| HDGCNInitialized embeddings=GloVe2021.06 | 89.12 | 70.37 | |
| TNet-LFSource=Zhang et al. (2019)2020.10 | 89.07 | 70.43 | |
| TNet-LFInitialized embeddings=GloVe2021.06 | 89.07 | 70.43 | |
| ASGCN-DGInitialized embeddings=GloVe2021.06 | 88.99 | 67.48 | |
| CPA-SA2026.04 | 88.92 | 72.43 | |
| RMN-P2026.04 | 88.75 | 71.54 | |
| SA-LSTM-PSource=Original Paper2020.10 | 88.7 | — | |
| ASGCNSource=Zhang et al. (2019)2020.10 | 88.69 | 66.64 | |
| ASGCN-DTInitialized embeddings=GloVe2021.06 | 88.69 | 66.64 | |
| HDGCNInitialized embeddings=BERT-base2021.06 | 87.99 | 71.28 | |
| ASGCNSource=Reproduced by authors, Runs=Average of 32020.10 | 87.71 | 70.29 | |
| AOAInitialized embeddings=GloVe2021.06 | 87.5 | 66.21 | |
| ASCNNSource=Zhang et al. (2019)2020.10 | 87.39 | 64.56 | |
| ASCNNInitialized embeddings=GloVe2021.06 | 87.39 | 64.56 | |
| MemNetSource=Zhang et al. (2019)2020.10 | 85.44 | 65.99 | |
| IANSource=Zhang et al. (2019)2020.10 | 84.74 | 55.21 | |
| IAN2026.04 | 83.79 | 58.62 | |
| PGCNN2026.04 | 81.34 | 58.23 |