Ambient Dataloops: Generative Models for Dataset Refinement
cs.LG, cs.AI
Submitted: 2026-01-21
Updated: 2026-09-16
Comments: 29 pages, 11 figures, 17 tables. Accepted at ICML 2026
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Iterated Denoising Energy Matching for Sampling from Boltzmann Densities
- Self-Improving Diffusion Models with Synthetic Data
- Training Diffusion Models with Reinforcement Learning
- Subadditivity of the log-Sobolev constant on convolutions
- Ambient Diffusion Omni: Training Good Models with Bad Data
- Iterative Importance Fine-tuning of Diffusion Models
- Diffusion Models Beat GANs on Image Synthesis
- A Tale of Tails: Model Collapse as a Change of Scaling Laws
- Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control
- Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification
- Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences
- Self-Correcting Self-Consuming Loops for Generative Model Training
- Scaling Laws for Autoregressive Generative Modeling
- Classifier-Free Diffusion Guidance
- Training Compute-Optimal Large Language Models
- Scaling Laws for Neural Language Models
- Segment Anything
- Noise2Noise: Learning Image Restoration without Clean Data
- ADG: Ambient Diffusion-Guided Dataset Recovery for Corruption-Robust Offline Reinforcement Learning
- Diffusion Beats Autoregressive in Data-Constrained Settings
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: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks