Unverified50% confidenceSolutionExact time
Approaches that fall between frozen embeddings and full fine-tuning include fine-tuning only the last few layers, using parameter-efficient fine-tuning methods such as LoRA, or employing an adapter architecture.
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
UnverifiedThe parameter-efficient fine-tuning (PEFT) field includes methods such as LoRA, QLoRA, Prefix Tuning, Adapter, and IA³, with a shared design philosophy of freezing most parameters and training only a small number of newly added parameters.70% similarUnverifiedCommon fine-tuning methods include full parameter fine-tuning, LoRA (Low-Rank Adaptation), and parameter-efficient techniques such as QLoRA.67% similarUnverifiedThe representative work of soft prompt techniques is Prompt Tuning proposed by Lester et al. 2021, which adds trainable continuous vectors only at the front of the input sequence while keeping model parameters frozen.65% similarVerified混合专家架构MoE通过稀疏激活机制，推理时只激活模型总参数的1/8至1/463% similarUnverifiedUnsloth同时支持LoRA微调和全量微调两种方式63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/711030API
curl https://kongchang.com/api/v1/knowledge/claims/711030MCP
get_claim(id=711030)