Question Answering on PubMedQA 80/20 split (seed 42)
0Pathwise ViolationsCSA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CSAN (EVAL size)=800, Err (Base verifier-fail rate)=23.9%, alpha* (Target risk level)=0.202026.05 | 0 | 63.5 | 15.1 | |
| NEX-ConfN (EVAL size)=800, Err (Base verifier-fail rate)=23.9%, alpha* (Target risk level)=0.202026.05 | 1 | — | — | |
| 5 online baselinesN (EVAL size)=800, Err (Base verifier-fail rate)=23.9%, alpha* (Target risk level)=0.202026.05 | 10 | — | — | |
| ConfFactN (EVAL size)=800, Err (Base verifier-fail rate)=23.9%, alpha* (Target risk level)=0.202026.05 | — | — | 7.6 |