Unverified50% confidenceFactExact time
The risk of pseudo-labels lies in confirmation bias—erroneous predictions get reinforced; therefore in practice novelty detection (e.g., Mahalanobis distance) is usually combined to judge whether the distribution has drifted, and pseudo-label updates are enabled only when the drift is significant and prediction confidence is high.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraditional pseudo-labeling suffers from confirmation bias: repeatedly using pseudo-labels from the same model causes biases to accumulate and amplify with each iteration.80% similarUnverifiedUnder domain mismatch conditions, relying solely on confidence-based filtering is insufficient to guarantee pseudo-label quality, because models may also exhibit high confidence on incorrect samples.76% similarUnverifiedAnomaly detection methods are friendlier to unknown defect types but have a relatively high false positive rate; synthetic data can expand sample size but has a domain gap from the real defect distribution.72% similarUnverified预标注会引入锚定效应，标注员倾向于接受而非推翻已有预标注结果，即使存在微小错误，低质量预标注会系统性地将判断拉向错误方向72% similarUnverifiedThe detector called by halu_guard returns calibrated probability values and annotates error types such as fabricated facts or false citations.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1057120API
curl https://kongchang.com/api/v1/knowledge/claims/1057120MCP
get_claim(id=1057120)