Unverified50% confidenceSolutionExact time
在监督微调（SFT）中，数据的分布往往比数据的数量更关键，精准的数据补充胜过盲目扩充数据集
1
Sources
50%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRMSNorm相比LayerNorm对数值范围约束更强，输出分布更集中，减少量化时的溢出风险69% similarUnverifiedFID (Fréchet Inception Distance) evaluates generation quality by comparing the statistical distribution differences between real and generated data in the feature space of a deep neural network, with lower values indicating closer resemblance to real data.65% similarUnverified在垂直领域（如法律、医疗）表现不佳时，通过收集高质量问答数据进行微调往往比无限优化Prompt更有效且更可持续65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/795902API
curl https://kongchang.com/api/v1/knowledge/claims/795902MCP
get_claim(id=795902)