List your company

Multi-task Language Understanding on MMLU-Pro n=8312 (test)

100ActivationCRC
67.86476.20784.5592.893Jun 28, 2026

Evaluation results

MethodLinks
CRCOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0610079.920.1531
Isotonic Conf.Operating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0698.384.415.4401.3
Selective Pred.Operating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0697.48514.6381.3
Calibrated LearnedOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0697.384.515.1401.3
Learned StopperOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0696.683.616.4431.3
ConsensusOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0694.584.515.5411.7
OracleOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0686100001.5
Adaptive Local k-NN Certification (kNN no bias)Operating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0684.390.48.1211.6
Adaptive Local k-NN CertificationOperating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0671.193.94.4122.2
Confidence Thr.Operating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0669.884.515.5412.4
Adaptive Local k-NN Certification (Final-round)Operating Budget (T)=4, Scaling parameter (λ*)=2.4, Target budget (β)=.38, Training error (eT)=.162026.0669.194.43.9104