Sequential Model Editing on CounterFact (test)
96.2ReliabilityDAFNet
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DAFNetBackbone=GPT-J (6B), # Editing=102024.05 | 96.2 | 65.8 | 85.2 | 82.4 | |
| DAFNetBackbone=LLAMA2 (7B), # Editing=12024.05 | 92 | 86.7 | 94.3 | 91 | |
| DAFNetBackbone=LLAMA2 (7B), # Editing=102024.05 | 87.3 | 59.6 | 85.9 | 77.6 | |
| DAFNetBackbone=GPT-J (6B), # Editing=1002024.05 | 81.8 | 40.3 | 87.3 | 69.8 | |
| DAFNetBackbone=LLAMA2 (7B), # Editing=1002024.05 | 72.8 | 41.5 | 76.4 | 63.6 |