Unverified50% confidenceFactExact time
Zero-training prediction still requires labels to be provided as in-context examples, but the labels do not participate in gradient updates.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCalibration must be fit on held-out labeled data, not on the training set, otherwise it will overfit the score distribution.70% similarUnverified作者选择不做标签规范化以保留零样本泛化能力，但代价是9000个单例导致大量标签缺乏监督信号且训练不稳定64% similarUnverifiedUnder this framework, few-shot learning corresponds to the constraint harness restricting the use of only a very small number of labeled samples, while the verifier still uses the standard test set.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1059306API
curl https://kongchang.com/api/v1/knowledge/claims/1059306MCP
get_claim(id=1059306)