Unverified50% confidenceFactExact time
As the number of model layers increases, the parameter share of Attention and FFN rises significantly while the share of Embedding gradually decreases
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified过度膨胀的上下文会稀释模型注意力、抬高推理成本，甚至引入噪声导致幻觉69% similarUnverifiedCurrent frontier models face attention decay when handling long-range dependencies, with the ability to maintain consistency of early-established settings declining significantly as context length increases.69% similarUnverifiedContext pollution causes the model's attention to become scattered, response quality to decline, and costs to spiral out of control.69% similarUnverified数据规模的增长并不等同于有效学习信号的增长，在模型能力提升到一定阈值后简单堆砌样本的收益会急剧下降68% similarUnverifiedHaving too many tokens in a prompt dilutes the attention mechanism's ability to focus, reducing the actual influence of each description.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/904294API
curl https://kongchang.com/api/v1/knowledge/claims/904294MCP
get_claim(id=904294)