Unverified50% confidenceConceptExact time
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.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn-the-wild benchmarks refer to deepfake datasets collected or generated in uncontrolled environments, characterized by variable lighting, cluttered backgrounds, free speaker poses, possibly moving cameras, and audio often accompanied by noise and reverberation.76% similarUnverifiedHard negative samples (audio similar to the target keyword but that should not trigger, such as homophones and similar phrases) can force the model to focus on truly discriminative features, thereby reducing false activations in real environments, and are more effective than simply increasing the amount of positive samples.75% similarVerifiedThe three main causes of hallucinations in large models are training data noise, the probability-first mechanism, and insufficient data timeliness.74% similarVerifiedIf samples are too easy the model has nothing to learn; if they are too difficult, noisy learning signals may interfere with training; the most effective learning occurs in the region slightly above current capability but still within reach.74% similarUnverifiedThe expected error of a supervised learning model on unseen data can be decomposed into three components: bias², variance, and irreducible noise74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028178API
curl https://kongchang.com/api/v1/knowledge/claims/1028178MCP
get_claim(id=1028178)