mlsanityData to 14 Oct 2026

Value Gradient Sampler: Learning Invariant Value Functions for Equivariant Diffusion Sampling

arXiv 2502.13280 ↗First posted 18 Feb 2025cs.LGcited by 0

On the map: Diffusion models → Diffusion and flow policies for robotics

Abstract (arXiv)

We propose the Value Gradient Sampler (VGS), a diffusion sampler parameterized by value functions. VGS generates samples from an unnormalized target density (i.e., energy) by evolving randomly initialized particles along the gradient of the value function. In many sampling problems where the target density exhibits invariant symmetries, value functions provide a novel approach to leveraging invariant networks for sampling by inducing an equivariant gradient flow, without requiring more complex equivariant networks. The value networks are trained via temporal difference learning, which supports off-policy training and other established reinforcement learning (RL) techniques. By combining advanced RL methods with efficient invariant networks, VGS achieves both the highest sample quality and the fastest sampling speed among our baselines on the 55-particle Lennard-Jones system.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 17 of its 52 references that are in the map. First posted on arXiv 18 Feb 2025.

Builds on

17 of 52 references.

References · 17

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 746
  3. 2020Equivariant Flows: Exact Likelihood Generative Learning for Symmetric Densitiescited by 250
  4. 2022e3nn: Euclidean Neural Networkscited by 174
  5. 2022An optimal control perspective on diffusion-based generative modelingcited by 147
  6. 2023Equivariant flow matchingcited by 135
  7. 2021Annealed Flow Transport Monte Carlocited by 90
  8. 2022Continual Repeated Annealed Flow Transport Monte Carlocited by 66
  9. 2019Equivariant Flows: sampling configurations for multi-body systems with symmetric energiescited by 57
  10. 2023Diffusion Generative Flow Samplers: Improving learning signals through partial trajectory optimizationcited by 57
  11. 2024Beyond ELBOs: A Large-Scale Evaluation of Variational Methods for Samplingcited by 56
  12. 2024Sequential Controlled Langevin Diffusionscited by 48
5 more
  1. 2024Training Neural Samplers with Reverse Diffusive KL Divergencecited by 21
  2. 2024Dynamical Measure Transport and Neural PDE Solvers for Samplingcited by 20
  3. 2024Energy based diffusion generator for efficient sampling of Boltzmann distributionscited by 7
  4. 2024Diffusion-PINN Samplercited by 7
  5. 2022Proposal of a Score Based Approach to Sampling Using Monte Carlo Estimation of Score and Oracle Access to Target Densitycited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025On scalable and efficient training of diffusion samplers
  2. 2025VARD: Efficient and Dense Fine-Tuning for Diffusion Models with Value-based RL
  3. 2025Symmetry-Aware Steering of Equivariant Diffusion Policies: Benefits and Limits
  4. 2025Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics
  5. 2023Towards Controllable Diffusion Models via Reward-Guided Exploration
  6. 2026Learning To Sample From Diffusion Models Via Inverse Reinforcement Learning
  7. 2025From discrete-time policies to continuous-time diffusion samplers: Asymptotic equivalences and faster training
  8. 2026A unified perspective on fine-tuning and sampling with diffusion and flow models