科研动态

闭源实验室RL细节外泄

Nice thread, some RL details used by closed labs slowly leaking out.

Nathan Lambert2026-10-25 19:19未入选

阅读原文

摘要

好帖，闭源实验室用的一些RL细节正慢慢泄露出来。

摘要由模型根据原文撰写，以原文为准。

正文 · 原文

Nice thread, some RL details used by closed labs slowly leaking out.

引用Martin Marek@mrtnm

Tinker's new StableReinforce objective is Reinforce + Score Centering. Here's how I reverse-engineered it 🧵

在 X 查看被引用的帖子

来源：Nathan Lambert · x.com

引用原文

natolambert. Nice thread, some RL details used by closed labs slowly leaking out.[EB/OL]. (2026-10-25)[访问日期]. https://x.com/natolambert/status/2114315936953925952.