Unverified50% confidenceFactExact time
Transformer的因果注意力掩码保证每个位置只能看到历史信息，与金融预测不能使用未来数据的原则天然契合
1
Sources
50%
Confidence
Long-term
Relevance
7/23/2026
First Seen
Sources
Kronos金融基础模型：用AI读懂K线市场语言
Related Claims
UnverifiedTransformer的知识高度分布式编码，一个事实散布在数百万参数中，使增量更新难以精准定位存储特定旧知识的参数66% similarUnverifiedTransformer由Vaswani等人于2017年提出，其知识以高度分布式方式编码，导致增量更新时无法精准定位存储特定知识的参数64% similarUnverified基准过拟合和数据污染会导致模型通过记忆而非真正理解获得高分，使公开基准分数可信度存疑62% similarUnverifiedEven if a model supports a context window of 128K or longer, stuffing all historical information into the prompt does not guarantee the model will correctly utilize that information, as attention dilution can cause critical information to be overlooked.58% similarUnverified欺骗是模型在内部已表征正确信息的情况下仍输出误导性内容，是系统性偏离而非能力不足58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/598868API
curl https://kongchang.com/api/v1/knowledge/claims/598868MCP
get_claim(id=598868)