[KongchangAI]
ConceptPSV

Perturb-Select-Vote

A general training-free inference-time ensemble framework that improves model performance by generating perturbed predictors, selecting high-performing members on a validation set, and aggregating their predictions.

Timeline (last 90 days)

Oct 30

The researchers propose the Perturb-Select-Vote (PSV) framework, whose process consists of generating perturbed predictors, selecting high-performing members on a small validation set, and aggregating predictions.

Unverified50%

All Facts (1)

Source Articles