QGF performs test-time policy optimization for flow models in RL by guiding a behavior-cloned reference policy with value-function gradients, achieving strong results on high-dimensional offline RL benchmarks without additional policy training.
Q-learning with adjoint matching
13 Pith papers cite this work. Polarity classification is still indexing.
13
Pith papers citing it
abstract
We propose Q-learning with Adjoint Matching (QAM), a novel TD-based reinforcement learning (RL) algorithm that tackles a long-standing challenge in continuous-action RL: efficient optimization of an expressive diffusion or flow-matching policy with respect to a parameterized Q-function. Effective optimization requires exploiting the first-order information of the critic, but it is challenging to do so for flow or diffusion policies because direct gradient-based optimization via backpropagation through their multi-step denoising process is numerically unstable. Existing methods work around this either by only using the value and discarding the gradient information, or by relying on approximations that sacrifice policy expressivity or bias the learned policy. QAM sidesteps both of these challenges by leveraging adjoint matching, a recently proposed technique in generative modeling, which transforms the critic's action gradient to form a step-wise objective function that is free from unstable backpropagation, while providing an unbiased, expressive policy at the optimum. Combined with temporal-difference backup for critic learning, QAM consistently outperforms prior approaches on hard, sparse reward tasks in both offline and offline-to-online RL.
citation-role summary
background 1
method 1
citation-polarity summary
years
2026 13representative citing papers
RLDT fine-tunes pretrained flow-matching policies for continuous control by aligning them to a max-entropy RL transport field constructed via SVGD, using expected-target estimation for stable multi-step updates.
FAV aligns few-step generative models by amortizing SVGD updates from reward-tilted sampling into generator parameters via fixed-point regression, requiring only sample access, and shows outperformance on robotics tasks plus scaling on image generators.
VGF solves behavior-regularized RL by transporting particles from a reference distribution to the value-induced optimal policy via discrete value-guided gradient flow.
Advantage-weighted conditional optimal transport flow matching simultaneously steers flow policies toward high-value actions and straightens their integration paths, enabling 2-3 step inference while improving task success.
FlowDPG distills critic gradients into flow matching velocity fields to enable BPTT-free DDPG-style policy improvement and reports 92% success on a real-world dual-arm AirPods assembly task.
VAST learns a horizon-adaptive auxiliary value function and stitching policy to compose variable-length returns for improved offline policy optimization on long-horizon tasks.
Reversal Q-Learning (RQL) proposes reversing flows for virtual trajectories and bias-variance reduction in an expanded MDP to train flow policies, reporting best average performance on 50 simulated robotic tasks versus prior flow-based offline RL methods.
DiPOD stabilizes diffusion policy optimization by interleaving self-distillation with gradient updates via an on-policy ELBO regularizer, yielding more stable training and higher rewards than prior methods.
Q-VGM offline RL fine-tuning converts critic Q-gradients into residual velocity targets for flow-matching VLAs, raising LIBERO success from 75.0% to 92.5%.
LWD is a fleet-scale offline-to-online RL framework that continually improves pretrained VLA policies using autonomous rollouts and human interventions, reaching 95% average success on real-world manipulation tasks.
FASTER models multi-candidate denoising as an MDP and trains a value function to filter actions early, delivering the performance of full sampling at lower cost in diffusion RL policies.
MBDPO reformulates policy optimization as a diffusion process over searched trajectories in latent world models to reduce misalignment between search and value learning.
citing papers explorer
-
Test-Time Gradient Guidance of Flow Policies in Reinforcement Learning
QGF performs test-time policy optimization for flow models in RL by guiding a behavior-cloned reference policy with value-function gradients, achieving strong results on high-dimensional offline RL benchmarks without additional policy training.
-
Reinforcement Learning for Flow-Matching Policies with Density Transport
RLDT fine-tunes pretrained flow-matching policies for continuous control by aligning them to a max-entropy RL transport field constructed via SVGD, using expected-target estimation for stable multi-step updates.
-
Aligning Few-Step Generative Models by Amortizing Sample-based Variational Inference
FAV aligns few-step generative models by amortizing SVGD updates from reward-tilted sampling into generator parameters via fixed-point regression, requiring only sample access, and shows outperformance on robotics tasks plus scaling on image generators.
-
Reinforcement Learning via Value Gradient Flow
VGF solves behavior-regularized RL by transporting particles from a reference distribution to the value-induced optimal policy via discrete value-guided gradient flow.
-
Optimal Transport Q-Learning for Flow Policy Steering and Acceleration
Advantage-weighted conditional optimal transport flow matching simultaneously steers flow policies toward high-value actions and straightens their integration paths, enabling 2-3 step inference while improving task success.
-
FlowDPG: Deterministic Policy Gradient on Flow Matching Policies for Real-World Manipulation
FlowDPG distills critic gradients into flow matching velocity fields to enable BPTT-free DDPG-style policy improvement and reports 92% success on a real-world dual-arm AirPods assembly task.
-
Horizon Adaptive Offline Policy Learning via Value Stitching
VAST learns a horizon-adaptive auxiliary value function and stitching policy to compose variable-length returns for improved offline policy optimization on long-horizon tasks.
-
Reversal Q-Learning
Reversal Q-Learning (RQL) proposes reversing flows for virtual trajectories and bias-variance reduction in an expanded MDP to train flow policies, reporting best average performance on 50 simulated robotic tasks versus prior flow-based offline RL methods.
-
DiPOD: Diffusion Policy Optimization without Drifting Apart
DiPOD stabilizes diffusion policy optimization by interleaving self-distillation with gradient updates via an on-policy ELBO regularizer, yielding more stable training and higher rewards than prior methods.
-
Q-VGM: Q-Value-Gradient Matching for Off-Policy Reinforcement Learning of Flow-Matching VLA
Q-VGM offline RL fine-tuning converts critic Q-gradients into residual velocity targets for flow-matching VLAs, raising LIBERO success from 75.0% to 92.5%.
-
Learning While Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies
LWD is a fleet-scale offline-to-online RL framework that continually improves pretrained VLA policies using autonomous rollouts and human interventions, reaching 95% average success on real-world manipulation tasks.
-
FASTER: Value-Guided Sampling for Fast RL
FASTER models multi-candidate denoising as an MDP and trains a value function to filter actions early, delivering the performance of full sampling at lower cost in diffusion RL policies.
-
Scaling World-Model Reinforcement Learning Through Diffusion Policy Optimization
MBDPO reformulates policy optimization as a diffusion process over searched trajectories in latent world models to reduce misalignment between search and value learning.