Minority Evidence
In a multi-reader parallel verification architecture, minority evidence refers to marginal evidence flagged as relevant by only a single reader while overlooked by all others. It constitutes a low-consensus signal in ensemble learning frameworks, typically residing at semantic boundaries and difficult for standard relevance models to capture consistently.
Timeline (last 90 days)
Under the k=50 setting, at most only 2 out of 8 minority evidence items can survive.
Minority evidence refers to information marked as relevant by only a single reader while all other readers failed to notice it.
Aggressive truncation strategies disproportionately lose marginal but potentially critical minority evidence.
All Facts (3)
Under the k=50 setting, at most only 2 out of 8 minority evidence items can survive.
50%UnverifiedMinority evidence refers to information marked as relevant by only a single reader while all other readers failed to notice it.
50%UnverifiedAggressive truncation strategies disproportionately lose marginal but potentially critical minority evidence.
50%