False Negative
In contrastive learning, a semantically matching sample that is mistakenly treated as a negative example because of incomplete annotation. This forces the model to push apart semantically similar samples, distorting the embedding space, and is a common source of training noise in multimodal contrastive learning.
Timeline (last 90 days)
A reader described a check that is structurally incapable of failing: an automated mobile check hashed screenshots before and after a swipe, but when a video was playing on screen the hash changed every frame, causing the check to pass regardless of whether the swipe took effect.
A false positive is when a test reports failure but there is actually no bug, and a false negative is when a test reports pass but a bug actually exists.
False negatives are a special type of noise in contrastive learning: the dataset contains samples that semantically match the anchor sample highly but are not recorded as positives due to incomplete annotation, and when they participate in training as negatives they distort the embedding space.
The false negative problem systematically underestimates retrieval quality, rendering comparisons between different methods meaningless.
In a risk control scenario, a false positive means misclassifying a legitimate transaction as fraudulent and blocking it, while a false negative means misclassifying a fraudulent transaction as legitimate and allowing it through.
All Facts (5)
A false positive is when a test reports failure but there is actually no bug, and a false negative is when a test reports pass but a bug actually exists.
50%UnverifiedFalse negatives are a special type of noise in contrastive learning: the dataset contains samples that semantically match the anchor sample highly but are not recorded as positives due to incomplete annotation, and when they participate in training as negatives they distort the embedding space.
50%UnverifiedThe false negative problem systematically underestimates retrieval quality, rendering comparisons between different methods meaningless.
50%UnverifiedIn a risk control scenario, a false positive means misclassifying a legitimate transaction as fraudulent and blocking it, while a false negative means misclassifying a fraudulent transaction as legitimate and allowing it through.
50%UnverifiedA reader described a check that is structurally incapable of failing: an automated mobile check hashed screenshots before and after a swipe, but when a video was playing on screen the hash changed every frame, causing the check to pass regardless of whether the swipe took effect.
50%