The paper develops an action-operator semantics for molecular diffusion models that supports auditable free-energy estimation from endpoint data.
ArXiv abs/2506.17139
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
years
2026 5representative citing papers
SCALLOP replaces Hutchinson's trace estimator with a scalable, vectorized likelihood distillation objective for F2D2 flow maps, cutting training variance and time while improving performance on molecular Boltzmann generators and image data.
SKMD adapts Stein variational gradient descent into molecular dynamics with asynchronous updates and global atomic descriptor kernels to acquire non-redundant training configurations while preserving the Boltzmann distribution, yielding higher MLIP accuracy with fewer samples than baselines.
A learned implicit bias in generative protein emulators accelerates state coverage up to 37x faster on Fast-Folding proteins while increasing diversity by 35%.
FES-FM learns a reduced flow-matching transport in collective-variable space to sample free energy surfaces, cutting per-sample generation cost while leaving full-space training cost unchanged.
citing papers explorer
-
Unsupervised Thermodynamics of Molecular Diffusion Models: Action-Operator Semantics and Auditable Free-Energy Readout
The paper develops an action-operator semantics for molecular diffusion models that supports auditable free-energy estimation from endpoint data.
-
Few-Step Boltzmann Generators via Scalable Likelihood Flow Maps
SCALLOP replaces Hutchinson's trace estimator with a scalable, vectorized likelihood distillation objective for F2D2 flow maps, cutting training variance and time while improving performance on molecular Boltzmann generators and image data.
-
Stein Kernelized Molecular Dynamics for Active Learning of Interatomic Potentials
SKMD adapts Stein variational gradient descent into molecular dynamics with asynchronous updates and global atomic descriptor kernels to acquire non-redundant training configurations while preserving the Boltzmann distribution, yielding higher MLIP accuracy with fewer samples than baselines.
-
Learning Implicit Bias in Generative Spaces for Accelerating Protein Dynamics Emulation
A learned implicit bias in generative protein emulators accelerates state coverage up to 37x faster on Fast-Folding proteins while increasing diversity by 35%.
-
FES-FM: Free Energy Surface Sampling via Reduced Flow Matching
FES-FM learns a reduced flow-matching transport in collective-variable space to sample free energy surfaces, cutting per-sample generation cost while leaving full-space training cost unchanged.