迈向AGI的扩展是否取决于System-1让System-2执行什么？
原标题：Paper page - Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?
AI 摘要
该论文提出「条件交互」框架：由轻量、具备语义感知的 System-1 预测器先评估候选交互的效用，再通过自适应阈值筛选，只让资源密集的 System-2 执行值得付出成本的交互，并用执行结果更新 System-1。作者在理想筛选与等执行成本假设下推导出执行计算量可降至约 1/ρ_u（ρ_u 为效用相关交互密度），并给出由预测误差与效用间隔界定的选择遗憾上界。论文据此提出核心假设：ρ_u 是可度量的扩展效率坐标，用「有效计算」而非总计算量描述扩展更合理。
正文节选
Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute? Abstract Co-designing a resource-heavy reasoning system with a lightweight predictive evaluator can reduce execution costs by filtering interactions, with efficiency governed by the density of useful interactions. Progress toward artificial general intelligence (AGI) has largely relied on scaling a resource-intensive executive, System-2, through larger models, more data and compute, longer reasoning, and more agents. We ask w