Google论文：声明式注意力机制降低LLM推理成本52%
New Google Deployment Paper shows instead of forcing an LLM to reread its entire…
长上下文推理是刚需，这篇论文给出了具体可落地的优化方案，显著降低算力消耗且精度损失可控，值得工程团队关注。
New Google Deployment Paper shows instead of forcing an LLM to reread its entire context for every token, let the model choose what to read
新的 Google 部署论文表明，与其强制 LLM 为每个 token 重读其整个上下文，不如让模型自行选择要读取的内容
cutting attention work by 52.0%
注意力计算量减少 52.0%
Today’s models often reread the entire stored context for every new token, even when only a small section matters.
当前模型在生成每个新 token 时，往往需要重读整个存储的上下文，即使其中只有一小部分相关。
Declarative Attention changes that: the model itself says which part of the context it needs, and the inference engine skips the rest.
声明式注意力（Declarative Attention）改变了这一现状：模型自身会指明它需要上下文的哪一部分，推理引擎则跳过其余部分。
No separate scorer is needed to search the whole context first.
无需先搜索整个上下文的独立评分器。
Across 15 long-context tasks, attention work fell by 52.0% on Gemma-4-31B and 31.1% on Qwen-3.6-27B, while accuracy dropped by 1.27 and 2.75 percentage points.
在 15 个长上下文任务中，Gemma-4-31B 的注意力计算量下降了 52.0%，Qwen-3.6-27B 下降了 31.1%，而准确率分别下降了 1.27 和 2.75 个百分点。
Larger models handled this trade-off better, suggesting the approach could improve further with training.
更大的模型更好地处理了这种权衡，表明该方法通过训练还有进一步提升的空间
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力