Local Ai
Positive-Only Drifting Policy Optimization
arXiv:2604.16519v1 Announce Type: new Abstract: In the field of online reinforcement learning (RL), traditional Gaussian policies and flow-based methods are often constrained by their unimodal express
arXiv:2604.16519v1 Announce Type: new Abstract: In the field of online reinforcement learning (RL), traditional Gaussian policies and flow-based methods are often constrained by their unimodal expressiveness, complex gradient clipping, or stringent trust-region requirements. Moreover, they all rely on post-hoc penalization of negative samples to correct erroneous actions. This paper introduces Positive-Only Drifting Policy Optimization (PODPO), a likelihood-free and gradient-clipping-free generative approach for online RL. By leveraging the drifting model, PODPO performs policy updates via advantage-weighted local contrastive drifting. Relying solely on positive-advantage samples, it elegantly steers actions toward high-return regions while exploiting the inherent local smoothness of the generative model to enable proactive error prevention. In doing so, PODPO opens a promising new pathway for generative policy learning in online settings.
Related
- Bayesian Ego-graph Inference for Networked Multi-Agent Reinforcement Learning
- LoDAdaC: a unified local training-based decentralized framework with adaptive gradients and compressed communication
- TempoNet: Slack-Quantized Transformer-Guided Reinforcement Scheduler for Adaptive Deadline-Centric Real-Time Dispatchs
- Beyond Single-Model Optimization: Preserving Plasticity in Continual Reinforcement Learning
Source: arXiv cs.LG | 2026-04-21