Unverified50% confidenceTradeoffExact time
Pure tournament selection (38.0) performs worse than the best single model (33.8), indicating that per-configuration selection suffers from overfitting the backtest.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在评估阶段对候选关系标签进行随机打乱后，该模型的F1从37%骤降到14%67% similarUnverified修复前模型每约29次下一token预测中约有1次的选择与数学上应有结果不同64% similarUnverifiedIn the most difficult single-concept quantitative category, most models' accuracy falls below 30%64% similarUnverifiedDixon-Coles模型在双泊松模型基础上引入了低比分修正项，弥补了0:0和1:0等低分比赛中泊松假设的偏差63% similarUnverified对比学习中的推远操作只要求负样本相似度低于正样本，并不要求达到-1的极端值63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/948989API
curl https://kongchang.com/api/v1/knowledge/claims/948989MCP
get_claim(id=948989)