RAG Claim Verification on FEVER (held-out test)
39.37AccuracyFedMark-FM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FedMark-FMGenerator=google/flan-t5-small, Scenario=Prompt-injection poisoning, Evaluation Mode=Held-out serving2026.07 | 39.37 | 27.18 | 0 | 0 | |
| VolumeGenerator=google/flan-t5-small, Scenario=Prompt-injection poisoning, Evaluation Mode=Held-out serving2026.07 | 31.87 | 23.62 | 100 | 1 | |
| Leave-one-outGenerator=google/flan-t5-small, Scenario=Prompt-injection poisoning, Evaluation Mode=Held-out serving2026.07 | 31.25 | 22.69 | 100 | 3.6 | |
| FL-ShapleyGenerator=google/flan-t5-small, Scenario=Prompt-injection poisoning, Evaluation Mode=Held-out serving2026.07 | 31.25 | 21.02 | 80 | 2.6 |