
Conceptually, EFM mirrors the inference procedure in Deep Equilibrium Models (@zicokolter @shaojieb) where a neural network is iterated to convergence at test time to find a fixed point.
Two paths, one equilibrium. ⚖️
English
Samir dar
18 posts

@Samir_Darouich
PhD student in computational chemistry and machine learning


Excited to share Equilibrium Matching (EqM)! EqM simplifies and outperforms flow matching, enabling strong generative performance of FID 1.96 on ImageNet 256x256. EqM learns a single static EBM landscape for generation, enabling a simple gradient-based generation procedure.




