Unverified50% confidenceFactExact time
Adversarial examples often lie at the edge of or outside ray coverage, while mislabeled samples may be covered within the expanded regions of training samples, causing the model to "confidently" give wrong predictions.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedSelf-supervised anomaly detection trains a reconstruction model using only normal samples, and at inference time treats regions with high reconstruction error as anomalies, naturally suited for scenarios with severe positive-negative sample imbalance.79% similarUnverifiedAutoregressive models suffer from the exposure bias problem, where the model sees real prefixes during training but faces its own erroneous generated prefixes during inference, leading to error accumulation.76% similarVerifiedSome models carry the risk of over-optimization toward test sets, causing leaderboard scores to deviate from real-world performance (benchmark saturation).75% similarUnverifiedThe neural expansion framework explains OOD samples as inputs that fall outside the ray coverage of training samples, so the model's predictions on them lack stability and may exhibit high uncertainty or random predictions.75% similarUnverifiedIf multiple models share similar training biases, multi-model adversarial debate may only amount to echo-chamber agreement and may fail to genuinely expose errors (model homogenization bias)74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1065869API
curl https://kongchang.com/api/v1/knowledge/claims/1065869MCP
get_claim(id=1065869)