Training Data Attribution on FAKEWIKI RolePlay
62.5Recall@10SCORINGMODEL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SCORINGMODELEvaluation Model Scope=Averaged across nine target LLMs2026.05 | 62.5 | 19.6 | 12.4 | |
| STEERFUSEEvaluation Model Scope=Averaged across nine target LLMs2026.05 | 50.1 | — | — | |
| Best baselineSelection Strategy=Selected from eleven retrieval baselines separately for each model, Evaluation Model Scope=Averaged across nine target LLMs2026.05 | 42.9 | — | — |