Noam Brown· @polynoamial · X·· 2026-06-09AI 评分68
AI 导读
自 @OpenAI o1 以来,我们就知道 LLM 测试时计算缩放。 然而两年后,实验室仍在报告模型的标量评测;安全组织仍对某个脚手架通过 100 倍推理表现更好感到惊讶;而 RSP 在决定关键阈值时仍忽略推理预算。
正文 · 原文
We've known about LLM test-time compute scaling since @OpenAI o1.
Yet 2 years later labs still report scalar evals for models; safety orgs are still surprised when a scaffold does better via 100x inference; and RSPs still ignore inference budget when deciding critical thresholds.
http://x.com/i/article/2057694226981257216在 X 查看被引用的帖子
来源:Noam Brown · x.com