Noise in Diffusion Models Is a Learnable Input
cs.LG, stat.ML
Submitted: 2026-08-03
Updated: 2026-09-08
Code: https://github.com/happyflatfish/prng-diffusion-learnability
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Couple to Control: Joint Initial Noise Design in Diffusion Models
- The FID Lottery: Quantifying Hidden Randomness in Generative-Model Evaluation
- DiffusionHijack: Supply-Chain PRNG Backdoor Attack on Diffusion Models and Quantum Random Number Defense
- Learning from Pseudo-Randomness With an Artificial Neural Network - Does God Play Pseudo-Dice?
- (How) Can Transformers Predict Pseudo-Random Numbers?
- Learning Pseudorandom Numbers with Transformers: Permuted Congruential Generators, Curricula, and Interpretability
- Non Gaussian Denoising Diffusion Models
- Score-based Denoising Diffusion with Non-Isotropic Gaussian Noise Models
- Deep Learning Scaling is Predictable, Empirically
- Auto-Encoding Variational Bayes
- Statistical Quality and Reproducibility of Pseudorandom Number Generators in Machine Learning technologies
- The influence of the random numbers quality on the results in stochastic simulations and machine learning
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks