Unverified50% confidenceBenchmarkExact time
在MS MARCO上随机批次训练损失快速收敛但泛化性能落后于Hobbit构造的难批次，后者训练损失更高但检索准确率大幅领先
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified在稠密检索中，批内负样本训练要求批次包含有信息量的负样本，随机采样会导致训练损失快速收敛但泛化性能差74% similarVerifiedOverfitting refers to the phenomenon where a model performs excellently on training data but its performance drops significantly on unseen test data.71% similarUnverifiedOn convergent tasks, there is a clear performance saturation phenomenon after models reach a certain scale, where continued increases in parameter count or training data yield rapidly diminishing marginal contributions to accuracy.69% similarUnverified仅用成功轨迹训练的模型往往存在分布外泛化的严重缺陷，遇到工具返回错误时便丧失纠偏能力68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/527096API
curl https://kongchang.com/api/v1/knowledge/claims/527096MCP
get_claim(id=527096)