Skip to content
Trending storyDeveloping

Merve Noyan says domain post-training of open models is a company's biggest moat

1 article1 sourcesince Oct 17Last article 1h ago ·

Overview

AISummary of 1 article

Hugging Face's Merve Noyan argues that a company's biggest moat is its ability to post-train open models on its own domain data using open RL environments, and says both are costly to build.

In an X post she says companies not doing this are missing out, and she cites NVIDIA's Nemotron 3.5 Super VL, which partners have adapted through post-training. The report contains only Noyan's claim and does not measure any results from such post-training.

Written by AI from the articles below · updated Oct 17, 7:04 AM ET

Check the sources:

Latest developmentOct 17, 6:48 AM ET
Merve Noyan says post-training open models on domain data is a key moat

Article timeline

The articles in this story. Times are ET.

Oct 17
  1. merveX
    Merve Noyan says post-training open models on domain data is a key moat

    AIHugging Face's Merve Noyan argues that a company's biggest moat is the ability to post-train open models on its own domain data using open RL environments, both costly to build. She says companies not doing this are missing out. Her post cites NVIDIA's Nemotron 3.5 Super VL, which partners have adapted through post-training.

Heat trend

Not enough continuous observations to show a trend yet.