Unverified50% confidenceFactExact time
灾难性遗忘是指模型在学习新知识时可能覆盖已有知识的现象
1
Sources
50%
Confidence
Long-term
Relevance
8/23/2026
First Seen
Sources
Related Entities
Related Claims
Unverified激进的全量微调或高强度DreamBooth训练可能通过灾难性遗忘破坏模型原有的角色知识74% similarUnverifiedCLAUDE.md选择外部记忆而非修改模型参数，规避了持续学习中的灾难性遗忘问题68% similarVerified根据情境学习理论，知识在其被应用的具体情境中习得时，记忆编码更深入、迁移能力更强68% similarUnverified研究表明真正的知识内化依赖于主动提取练习（Active Recall）和间隔重复（Spaced Repetition）67% similarUnverified微调既能注入有用领域知识，也可能无意中扭曲模型原有正确认知，是一把双刃剑66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/789967API
curl https://kongchang.com/api/v1/knowledge/claims/789967MCP
get_claim(id=789967)