Unverified50% confidenceBenchmarkExact time
At the actual repetition levels of ordinary text, five models ranging from 1B to 13B parameters carry only a faint trace of what they have seen, with a rank correlation coefficient of approximately -0.08.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified实验数据显示，经过5-10轮迭代训练后，模型在低频词汇和长尾知识上的覆盖率下降超过40%，而输出文本在表面流畅性上几乎没有可察觉的退化67% similarUnverified该测试样本规模相对有限（5个模型、13个代码库），结论有待更大规模验证66% similarUnverified少样本学习在提示词中附带少量输入-输出示例，通常1-5个66% similarUnverifiedFew-shot提示通过提供2-5个示例让模型理解模式，比纯文字描述准确率通常高出20%-40%66% similarUnverifiedThe --sref weight value has a maximum of 10; in this case it is set to 7, and the higher the value, the closer the generated result is to the reference style.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/907911API
curl https://kongchang.com/api/v1/knowledge/claims/907911MCP
get_claim(id=907911)