Precision@K estimation on ESCI
0.29BiasPPI + Claude 3 Haiku
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PPI + Claude 3 Haikun_gold=30, N_unlabeled=60,000, LLM_judge=Claude 3 Haiku2026.06 | 0.29 | 3.86 | 79 | |
| PPI + Claude 3 Sonnetn_gold=30, N_unlabeled=60,000, LLM_judge=Claude 3 Sonnet2026.06 | 0.7 | 3.5 | 946 | |
| Gold onlyn_gold=302026.06 | 1.04 | 4.45 | — |