Unverified50% confidenceFactExact time
Existing research shows that large language models exhibit a clear memory preference when parametric knowledge conflicts with retrieved evidence, tending to rely on knowledge acquired during training.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedAccording to situated learning theory, when knowledge is acquired in the specific context where it is applied, memory encoding is deeper and transfer ability is stronger.78% similarVerifiedIn-context learning was first observed in large language models, and its capability fundamentally stems from the diversity of pretraining data.78% 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）75% similarVerified当模型训练数据包含评测集题目或高度相似内容时，模型可能通过记忆而非真正理解获得高分73% similarUnverifiedParametric knowledge refers to the implicit memory compressed into the weight parameters of a language model during pre-training, while non-parametric knowledge refers to knowledge retrieved at inference time.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999283API
curl https://kongchang.com/api/v1/knowledge/claims/999283MCP
get_claim(id=999283)