Drift Q-Learning

arXiv cs.LG Papers

Summary

Proposes DriftQL, which combines a drift-based behavioral regularizer with critic-driven policy improvement for offline RL, outperforming diffusion and flow methods on D4RL and OGBench while maintaining simplicity and efficiency.

arXiv:2606.00350v1 Announce Type: new Abstract: Offline reinforcement learning requires improving a policy from fixed data while avoiding out-of-distribution actions with unreliable value estimates. Diffusion and flow policies handle this trade-off by modeling the behavior distribution to regularize the RL objective, but they require iterative denoising, solver integrations, and in more efficient variants, distillation or other approximations at inference. We propose DriftQL, which combines a drift-based behavioral regularizer with critic-driven policy improvement. The value signal biases the policy toward high-value regions of the data support, while attraction and repulsion together keep generated actions near the data and prevent collapse onto a single mode. DriftQL is implemented as a single network with a unified training objective and generates actions in a single forward pass. On D4RL and OGBench, DriftQL consistently outperforms diffusion and flow methods, advancing the state of the art. Under degraded data quality, where the baselines visibly struggle, DriftQL remains close to its clean-data performance, positioning it as a promising alternative to diffusion and flow-based methods while maintaining the simplicity and efficiency of deterministic approaches. Project page: https://driftql.github.io/
Original Article
View Cached Full Text

Cached at: 06/02/26, 03:41 PM

# Drift Q-Learning
Source: [https://arxiv.org/abs/2606.00350](https://arxiv.org/abs/2606.00350)
[View PDF](https://arxiv.org/pdf/2606.00350)

> Abstract:Offline reinforcement learning requires improving a policy from fixed data while avoiding out\-of\-distribution actions with unreliable value estimates\. Diffusion and flow policies handle this trade\-off by modeling the behavior distribution to regularize the RL objective, but they require iterative denoising, solver integrations, and in more efficient variants, distillation or other approximations at inference\. We propose DriftQL, which combines a drift\-based behavioral regularizer with critic\-driven policy improvement\. The value signal biases the policy toward high\-value regions of the data support, while attraction and repulsion together keep generated actions near the data and prevent collapse onto a single mode\. DriftQL is implemented as a single network with a unified training objective and generates actions in a single forward pass\. On D4RL and OGBench, DriftQL consistently outperforms diffusion and flow methods, advancing the state of the art\. Under degraded data quality, where the baselines visibly struggle, DriftQL remains close to its clean\-data performance, positioning it as a promising alternative to diffusion and flow\-based methods while maintaining the simplicity and efficiency of deterministic approaches\. Project page:[this https URL](https://driftql.github.io/)

## Submission history

From: Mohamad H Danesh \[[view email](https://arxiv.org/show-email/b71ab235/2606.00350)\] **\[v1\]**Fri, 29 May 2026 20:42:30 UTC \(1,995 KB\)

Similar Articles

Reversal Q-Learning

arXiv cs.LG

This paper proposes Reversal Q-Learning (RQL), an offline reinforcement learning algorithm that trains a flow policy using an expanded Markov decision process framework and techniques to enable off-policy RL without backpropagation through time. It achieves state-of-the-art performance on challenging simulated robotic tasks.

QPILOTS: Efficient Test-Time Q-Steering for Flow Policies

arXiv cs.LG

QPILOTS is a method that steers flow policies at inference time by using critic gradients projected from noisy intermediate states, achieving state-of-the-art performance on offline-to-online RL benchmarks and improving pretrained VLA models without modifying the base policy.