weight-space-learning

A collection of weight space learning including papers, codes, and datasets.

Official PyTorch Implementation for the "Recovering the Pre-Fine-Tuning Weights of Generative Models" paper (ICML 2024).

Official PyTorch Implementation for the "Unsupervised Model Tree Heritage Recovery" paper (ICLR 2025).

Awesome papers on weight-space learning

Code Repository for the ICML 2024 paper: "Towards Scalable and Versatile Weight Space Learning".

Code Repository for the NeurIPS 2022 paper: "Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights".

Official PyTorch Implementation for the "Learning on Model Weights using Tree Experts" paper (CVPR 2025).

Code Repository for the CVPR 2026 paper: "Learning Geospatial Representations from Models, not Data".

An official implementation of ProbeGen

AAAI'26 Oral: "WeightFlow: Learning Stochastic Dynamics via Evolving Weight of Neural Network"

Official implementation of Weight-Space Linear Recurrent Neural Networks (ICLR 2026)

ICML'26: Generative Adaptation of Dynamics to Environmental Shifts via Weight-space Diffusion

Measuring how much of the INR weight-space perception gap is parameter symmetry. Exact D∞≀Sₙ characterization for sine networks, an orbit-only intervention that isolates what the group alone costs, and a phasor-graded reader placing 3rd/3rd/2nd on the standard INR benchmarks. Pre-registered; every prediction scored.

[ICLR 2026] Weight Space Representation Learning on Diverse NeRF Architectures

Website for the ICLR 2025 Weight Space Learning workshop.

SANE backbone trained on multiple model zoos.