Subject Inference Attack on CounterFact
100Attack PerformanceKSTER
Evaluation Results
| Method | Links | |
|---|---|---|
| KSTERModel=GPT-J, Attack Setting=White-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 | |
| Gray-box AttackModel=GPT-J, Attack Setting=Gray-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 | |
| KSTERModel=Llama-3-Instruct, Attack Setting=White-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 | |
| Gray-box AttackModel=Llama-3-Instruct, Attack Setting=Gray-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 | |
| KSTERModel=Qwen-2.5-Instruct, Attack Setting=White-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 | |
| Gray-box AttackModel=Qwen-2.5-Instruct, Attack Setting=Gray-box, Editing Method=ROME, Edit Scale=Single-edit2026.02 | 100 |