8 papers
PoEM: Predicting RL Outcomes from Existing Policies
Kimia Hamidieh, Giannis Daras, Antonio Torralba
Foundation models are post-trained with reinforcement learning (RL) to maximize specific rewards, such as human alignment, correctness, or instruction following. This post-training…
Seeing Through the PRISM: Compound & Controllable Restoration of Scientific Images
Rupa Kurinchi-Vendhan, Pratyusha Sharma, Antonio Torralba +1
Scientific and environmental imagery often suffer from complex mixtures of noise related to the sensor and the environment. Existing restoration methods typically remove one degrad…
Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis
Hila Chefer, Patrick Esser, Dominik Lorenz +5
Strong semantic representations improve the convergence and generation quality of diffusion and flow models. Existing approaches largely rely on external models, which require sepa…
Ambient Dataloops: Generative Models for Dataset Refinement
Adrián Rodríguez-Muñoz, William Daspit, Adam Klivans +3
We propose Ambient Dataloops, an iterative framework for refining datasets that makes it easier for diffusion models to learn the underlying data distribution. Modern datasets cont…
Dataset Distillation for Pre-Trained Self-Supervised Vision Models
George Cazenavette, Antonio Torralba, Vincent Sitzmann
The task of dataset distillation aims to find a small set of synthetic images such that training a model on them reproduces the performance of the same model trained on a much larg…
DiffEM: Learning from Corrupted Data with Diffusion Models via Expectation Maximization
Danial Hosseintabar, Fan Chen, Giannis Daras +2
Diffusion models have emerged as powerful generative priors for high-dimensional inverse problems, yet learning them when only corrupted or noisy observations are available remains…
Ambient Diffusion Omni: Training Good Models with Bad Data
Giannis Daras, Adrian Rodriguez-Munoz, Adam Klivans +2
We show how to use low-quality, synthetic, and out-of-distribution images to improve the quality of a diffusion model. Typically, diffusion models are trained on curated datasets t…
Score Distillation Sampling for Audio: Source Separation, Synthesis, and Beyond
Jessie Richter-Powell, Antonio Torralba, Jonathan Lorraine
We introduce Audio-SDS, a generalization of Score Distillation Sampling (SDS) to text-conditioned audio diffusion models. While SDS was initially designed for text-to-3D generation…