Jun 13, 2024 · Our results indicate that the weight space of fine-tuned diffusion models can behave as an interpretable meta-latent space producing new models.

We investigate the space of weights spanned by a large collection of customized diffusion models. We populate this space by creating a dataset of over ...

Jun 13, 2024 · Our results indicate that the weight space of fine-tuned diffusion models behaves as an interpretable latent space of identities.

Finally, we show that enforcing weights to live in this space enables a diffusion model to learn a subject given a single image, even if it is out of ...

Jun 13, 2024 · It is found that inverting a single image into this space encodes a realistic identity into a model, even if the input image is out of ...

Our results indicate that the weight space of fine-tuned diffusion models can behave as an interpretable meta-latent space producing new models.

Jun 14, 2024 · We present weights2weights, a subspace in diffusion weights that behaves as an interpretable latent space over customized diffusion models.

Jun 13, 2024 · Title: Interpreting the Weight Space of Customized Diffusion Models Authors: Amil Dravid*, Yossi Gandelsman, Kuan-Chieh Wang, Rameen Abdal, ...

Jun 13, 2024 · First, as each point in the space corresponds to an identity, sampling a set ofweights from it results in a model encoding a novel identity.

Nov 25, 2024 · The paper explores interpreting the weight space of customized diffusion models, which are powerful generative models used for tasks like image ...