Unverified50% confidenceFactExact time
Several commonly used proxy metrics in the study (such as language fluency and instruction complexity) did not show a strong correlation with training utility.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedExternal scheduling at inference time and internally learned capabilities during training differ in generalization performance and controllability.78% similarUnverifiedWhether the simplex constraint limits the expressive freedom of embeddings, and the training stability under Total Variation similarity, still require more extensive experiments.74% similarUnverified模型能力高度依赖训练数据的分布，法律文本充足而空间推理题的文本化表示相对稀少73% similarVerifiedIf evaluation benchmarks are highly aligned with training data, surpassing them reflects overfitting-style specialization rather than genuine capability improvement.73% similarVerifiedAlignment training optimizes the model's statistical behavior in known scenarios rather than establishing formally verifiable security boundaries, and may fail when faced with rare instruction combinations or role-playing frameworks outside the training distribution.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102342API
curl https://kongchang.com/api/v1/knowledge/claims/1102342MCP
get_claim(id=1102342)