Researchers implemented fine tuning suppression in image generation models within a restricted domain, effectively creating a local minima that is difficult to escape. By constructing the loss function to have near-zero gradients during fine tuning, the model resists traditional gradient descent techniques, making it challenging to modify existing capabilities. This approach aims to make fine tuning more challenging than training from scratch, introducing a novel and intriguing concept in model development.

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.
App store bannerPlay store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode

Save any
moment

Hear something you like? Tap your headphones to save it with AI-generated key takeaways

Share
& Export

Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode