Aspect Sentiment Quad Prediction on ACOS Rest
67.4F1 ScoreS²IT-32B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| S²IT-32BMethod Paradigm=Generative-based, Model Scale=32B2026.04 | 67.4 | — | 66.6 | |
| S²IT-7BMethod Paradigm=Generative-based, Model Scale=7B2026.04 | 66.1 | — | 64.9 | |
| E4LTraining protocol=State-of-the-art2025.11 | 64.91 | 65.46 | 64.37 | |
| E4Lcollaborative=true2025.11 | 64.91 | — | — | |
| qwen2.5-32BMethod Paradigm=Generative-based, Model Scale=32B2026.04 | 64.2 | — | 65 | |
| MUL+Scorer&ReRank (AI)collaborative=true, refinement=Scorer & ReRank (AI)2025.11 | 63.63 | — | — | |
| qwen2.5-7BMethod Paradigm=Generative-based, Model Scale=7B2026.04 | 63.6 | — | 64.1 | |
| GAS+Scorer&ReRank (AI)collaborative=true, refinement=Scorer & ReRank (AI)2025.11 | 62.47 | — | — | |
| MvPTraining protocol=State-of-the-art2025.11 | 61.54 | — | — | |
| One-ASQPMethod Paradigm=Extractive-based2026.04 | 60.6 | — | 56.2 | |
| MvP + ATossMethod Paradigm=Generative-based2026.04 | 60.6 | — | 58.3 | |
| MULTraining protocol=State-of-the-art2025.11 | 60.53 | 61.22 | 59.87 | |
| DLO + ATossMethod Paradigm=Generative-based2026.04 | 60.5 | — | 59.9 | |
| DLOMethod Paradigm=Generative-based2026.04 | 59.9 | — | 59.8 | |
| ParaphraseMethod Paradigm=Generative-based2026.04 | 59.8 | — | 59.8 | |
| MvPMethod Paradigm=Generative-based2026.04 | 59.5 | — | 57.8 | |
| DLOTraining protocol=State-of-the-art2025.11 | 59.18 | 60.02 | 59.84 | |
| ILOTraining protocol=State-of-the-art2025.11 | 58.69 | 58.43 | 58.95 | |
| GASTraining protocol=State-of-the-art2025.11 | 58.63 | 59.81 | 57.51 | |
| Seq2PathTraining protocol=State-of-the-art2025.11 | 58.41 | — | — | |
| IVLSTraining protocol=State-of-the-art2025.11 | 55.25 | 60.46 | 51.14 | |
| Extract-ClassifyTraining protocol=State-of-the-art2025.11 | 44.61 | 38.54 | 52.96 | |
| Extract-ClassifyMethod Paradigm=Extractive-based2026.04 | 44.6 | — | 52.9 | |
| ChatGPTTraining protocol=Training-free, Shot count=few-shot2025.11 | 42.02 | 38.39 | 46.4 | |
| TAS-BERTMethod Paradigm=Extractive-based2026.04 | 33.5 | — | 46.3 |