Unverified50% confidenceOpinionExact time
Inconsistent label direction may stem from randomness in model outputs or uncertainty in evaluation criteria, posing a fundamental challenge to SafeRoute-MAS, which relies on labels for risk calibration.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/11/2026
First Seen
Valid until: 2/9/2027
Sources
Related Entities
Related Claims
UnverifiedThe 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.73% 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.71% similarUnverifiedDecision models are less reliable than LLMs as hard-label detectors of implanted false sources, but their output probabilities can still distinguish implanted sources from clean sources.70% similarUnverifiedSFT forces the model to incorporate incorrect pseudo-labels into the policy's support set, destroying reward variance and causing the model to lose the ability to distinguish right from wrong.70% similarUnverifiedPlacing the exact enum value immediately next to the natural-language label in the prompt (e.g., "Status: At risk (corresponding enum value: risk)") can reduce the probability of the model copying the wrong label.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1158204API
curl https://kongchang.com/api/v1/knowledge/claims/1158204MCP
get_claim(id=1158204)