@openai: Let's talk about evals. We're always looking for better ways to measure and forecast model progress, especially as benchmarks get saturated or gamed.

2026-07-28 18:34:47

2026-07-28 03:36:24

2026-07-27 20:40:01

2026-07-27 20:57:25

2026-07-27 20:26:40

2026-07-27 20:28:12

2026-07-28 02:04:26

2026-07-27 20:26:14

2026-07-27 20:29:52

2026-07-28 06:44:32

2026-07-27 20:37:28

2026-07-27 20:31:36

To see more videos from user @openai, please go to the Tikwm homepage.

@openai: Let's talk about evals. We're always looking for better ways to measure and forecast model progress, especially as benchmarks get saturated or gamed.

@openai: Let's talk about evals. We're always looking for better ways to measure and forecast model progress, especially as benchmarks get saturated or gamed.

OpenAI

Open In TikTok:

Region: US

Monday 27 July 2026 20:23:43 GMT

Music

Download

Comments

Eli Walther :

ChatGPT models have been really good at science so far

amackintosh 🇨🇦 :

I get so excited when I think about doing research with ai. My dream is to not have to work but choose to sit at the computer for long hours mining ai for new knowledge

unknownmoira333 :

That is so cool

FuadRocket🚀 :

Open Ai...i can not buy gpt pro from ethiopia...what should i do

buv76 :

the first

￴￴ :

First give free pro

edwintorres6381 :

se libreeeeee

🍒~evelyn~ fan de dayana~🍒 :

one hiiii

simply_mik4 :

Early

Nelxon :

😎😎😎

deux 🇵🇸🇧🇪🇲🇦 :

🔥🔥🔥

Cheick.dev👨‍💻🇲🇱 :

@Mia K. c'est ta soeur ?

Other Videos

About

Legal