Sentiment Editing on ConvSent
87.25Success Rate (1K Edits)CoT2Edit
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CoT2EditBackbone Model=Llama3-8B2026.04 | 87.25 | 86.18 | 85.33 | 85.27 | 85.2 | |
| CoT2EditBackbone Model=Falcon-3-10B2026.04 | 84.32 | 83.01 | 82.82 | 80.68 | 79.08 | |
| LTEBackbone Model=Llama3-8B2026.04 | 80.49 | 78.83 | 76.11 | 73.37 | 72.78 | |
| EditCoTBackbone Model=Falcon-3-10B2026.04 | 79.15 | 78.81 | 76.12 | 75.27 | 74.51 | |
| EditCoTBackbone Model=Llama3-8B2026.04 | 78.34 | 76.79 | 76.43 | 75.11 | 74.91 | |
| AlphaEditBackbone Model=Falcon-3-10B2026.04 | 78.07 | 77.75 | 74.63 | 72.18 | 70.7 | |
| LTEBackbone Model=Falcon-3-10B2026.04 | 77.37 | 76.15 | 75.76 | 72.62 | 71.03 | |
| AlphaEditBackbone Model=Llama3-8B2026.04 | 76.13 | 70.53 | 63.28 | 55.38 | 53.11 | |
| PMETBackbone Model=Falcon-3-10B2026.04 | 71.18 | 70.03 | 69.67 | 68.13 | 66.55 | |
| PMETBackbone Model=Llama3-8B2026.04 | 67.13 | 65.11 | 63.27 | 61.15 | 60.58 | |
| MEMITBackbone Model=Falcon-3-10B2026.04 | 61.48 | 57.72 | 53.17 | 50.58 | 47.36 | |
| MEMITBackbone Model=Llama3-8B2026.04 | 58.72 | 54.54 | 50.66 | 46.9 | 43.06 | |
| IKEBackbone Model=Llama3-8B2026.04 | 41.19 | 38.32 | 36.2 | 32.37 | 29.33 | |
| IKEBackbone Model=Falcon-3-10B2026.04 | 40.75 | 36.53 | 33.45 | 31.08 | 27.99 |