Unverified60% confidenceSolutionExact time
During fine-tuning, hard negatives (e.g., positions in the same field but at a different seniority level) should be used instead of random negatives to obtain maximum gains.
2
Sources
60%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedStratified sampling and inverse probability weighting are common statistical methods when evaluating rare events, restoring the true production distribution by assigning weights to samples, but they cannot make up for the insufficiency of the sample size itself.76% similarUnverifiedThe impact of data quality on fine-tuning results is often far greater than that of hyperparameter tuning, and this is especially critical in scenarios with limited sample sizes.76% similarUnverifiedUsing the median rather than the mean is to avoid disturbance of the metric by long-tail samples and to better reflect the handling experience of typical issues.75% similarUnverified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% similarUnverifiedData engineering also includes sample deduplication (to avoid training set leakage), difficulty stratification (to prevent simple samples from dominating the gradient), and quality filtering (to remove low-quality or harmful content).73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/827336API
curl https://kongchang.com/api/v1/knowledge/claims/827336MCP
get_claim(id=827336)