First Squawk(@FirstSquawk):OPENAI SAYS THAT ON AUGUST 28 IT RESTARTED A LARGE FRONTIER REINFORCEMENT LEARNING RUN THAT WAS PREVIOUSLY PAUSED, AND THAT FOR GPT-5.6 IT IMPROVED THE ROBUSTNESS OF ITS SYSTEM-LEVEL STACK BY ADDING ACTIVATION CLASSIFIERS AND IMPROVING COVERAGE OVER UNIVERSAL JAILBREAKS.

First Squawk

@FirstSquawk

Firstsquawk Is a Global Financial (News) Squawk Service Specialising In Financial Markets, Macro, Geopolitics,politics

加入 July 2015

251 正在關注 560.5K 粉絲

First Squawk@FirstSquawk

2026.09.01 20:04

OPENAI SAYS THAT ON AUGUST 28 IT RESTARTED A LARGE FRONTIER REINFORCEMENT LEARNING RUN THAT WAS PREVIOUSLY PAUSED, AND THAT FOR GPT-5.6 IT IMPROVED THE ROBUSTNESS OF ITS SYSTEM-LEVEL STACK BY ADDING ACTIVATION CLASSIFIERS AND IMPROVING COVERAGE OVER UNIVERSAL JAILBREAKS.