Unverified50% confidenceFactExact time
Taking d=5 and τ=10 as an example, a trainable proportion of about 4.1% captures 90% of the reduction in the approximation bound from P=0 to P=1.
1
Sources
50%
Confidence
Long-term
Relevance
10/23/2026
First Seen
Sources
Related Entities
Related Claims
Unverified若单一模型漏检率为 p，两个训练背景不同的模型同时漏检同一错误的概率近似降至 p²，20%漏检率理论上可降至 4%73% similarUnverifiedWhen sampling eight candidates without constraints from the model, 10.2% of inputs cannot produce any valid output under limited training data, and this proportion drops to 0.4% when using the full training data71% similarVerifiedWhen fine-tuning with LoRA, the actual number of parameters that need to be trained is only 0.1%-1% of the original.70% similarUnverified负采样将每步训练的复杂度从O(V)降低到O(k)，k是负样本数（通常取5-20）69% similarUnverifiedThe loss during training is calculated per token, and long samples contribute far more gradient than short samples, so a 10% share of rows can absolutely correspond to more than 50% of the token volume and gradient signal.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1088320API
curl https://kongchang.com/api/v1/knowledge/claims/1088320MCP
get_claim(id=1088320)