Unverified50% confidenceDecision RuleExact time
The choice between frozen embeddings and fine-tuning depends on the amount of labeled data, the computational budget, and the degree of distributional divergence between the target task and the pre-training task.
1
Sources
50%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
嵌入向量时代：数据科学从特征工程到表征选择的范式转移
Related Claims
Unverified参数量相似的情况下，训练数据质量分布的差异往往比架构差异更能决定特定场景的表现上限76% similarUnverified模型之间的差异源于训练数据规模和质量、参数数量、对齐策略（如RLHF）以及上下文窗口长度等核心设计选择72% similarUnverified训练-部署鸿沟指模型因训练与部署环境的数据分布偏移、软件依赖版本差异或硬件指令集不同而导致的性能下降71% similarUnverifiedThe Chinchilla law states that under a fixed compute budget, appropriately reducing model size and increasing training data volume often yields better performance.70% similarUnverified最大似然估计训练目标等价于最小化模型分布与训练数据经验分布之间的KL散度70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/711029API
curl https://kongchang.com/api/v1/knowledge/claims/711029MCP
get_claim(id=711029)