Backdoor Detection on SST-2
10TPRCSO(continuous)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CSO(continuous)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 10 | 1 | |
| CSO(continuous)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 10 | 1 | |
| CSO(continuous)Model=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 10 | 1 | |
| CSO(discrete)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 10 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 10 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 10 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 10 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 10 | 0 | |
| CSO(continuous)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 9 | 1 | |
| CSO(continuous)Model=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 9 | 1 | |
| CSO(discrete)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 9 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 9 | 0 | |
| CSO(discrete)Model=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 9 | 0 | |
| PICCOLOModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 9 | 0 | |
| PICCOLOModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 9 | 0 | |
| CSO(continuous)Model=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 8 | 1 | |
| CSO(continuous)Model=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 8 | 1 | |
| ALTModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 8 | — | |
| PICCOLOModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 8 | 0 | |
| PICCOLOModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 8 | 0 | |
| CSO(continuous)Model=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 7 | 1 | |
| ALTModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 7 | — | |
| ALTModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 6 | — | |
| PICCOLOModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 6 | 0 | |
| ALTModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 5 | — | |
| DBSModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 5 | 7 | |
| PICCOLOModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 5 | 0 | |
| PICCOLOModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 5 | 0 | |
| DBSModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 4 | 7 | |
| PICCOLOModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 4 | 0 | |
| ALTModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 3 | — | |
| DBSModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 3 | 7 | |
| DBSModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 3 | 7 | |
| ALTModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 2 | — | |
| ALTModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 2 | — | |
| DBSModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 2 | 7 | |
| DBSModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 2 | 7 | |
| CLIBEModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 2 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 2 | 10 | |
| ALTModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 1 | — | |
| DBSModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 1 | 7 | |
| DBSModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 1 | 7 | |
| CLIBEModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 1 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 1 | 10 | |
| MMModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=high2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=high2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=high2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 0 | 10 | |
| MMModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=high2026.06 | 0 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=clean label, Poisoning Rate=low2026.06 | 0 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=Tell me seriously, Attack Type=dirty label, Poisoning Rate=low2026.06 | 0 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=mb, Attack Type=clean label, Poisoning Rate=low2026.06 | 0 | 10 | |
| CLIBEModel=flan-t5-small, Trigger=mb, Attack Type=dirty label, Poisoning Rate=low2026.06 | 0 | 10 |