# Policy Gradient Methods

3 related articles

Subscribe to updates

arXiv Oct 23, 2026
Why curriculum learning pays off in reinforcement learning: sparse rewards can stall learning without one, while dense rewards gain little
Curriculum learning often helps more in reinforcement learning than in supervised learning, and a tractable target-interception model now explains why. With dense rewards, curricula bring only marginal gains; with sparse rewards, a curriculum can be required for learning to occur at all. When the agent must both minimize effort and accumulate reward, competing optima appear and can trap it in a quitting solution. Useful curricula do two things: they make rewards less sparse to speed learning, and they shape the reward landscape to steer agents away from bad local optima.
Curriculum learning often helps more in reinforcement learning than in supervised learning, and a tractable target-interception model now explains why. With dense rewards, curricula bring only marginal gains; with sparse rewards, a curriculum can be required for learning to occur at all. When the agent must both minimize effort and accumulate reward, competing optima appear and can trap it in a quitting solution. Useful curricula do two things: they make rewards less sparse to speed learning, and they shape the reward landscape to steer agents away from bad local optima.
Harvard University
John J. Vastola（First author）
Curriculum Learning Policy Gradient Methods Reward Sparsity
arXiv Oct 7, 2026
Unified Benchmark Plus RL Feature Selection: Learned Models Consistently Beat Fixed-Coefficient USGS Model for Post-Wildfire Debris Flow Prediction, with Drivers Varying Sharply by Region
The work builds a unified benchmark for data-driven post-wildfire debris flow (PFDF) prediction, uses an LLM to convert operational weather forecast narratives into structured meteorological indicators and augments features with terrain and soil-moisture context, and proposes a reinforcement-learning feature-selection framework that identifies key drivers by perturbing features until positive and negative events become indistinguishable; learned models consistently outperform the fixed-coefficient USGS logistic-regression baseline, augmented features further improve most model families, and storm variables dominate globally while feature importance varies substantially across regions.
The work builds a unified benchmark for data-driven post-wildfire debris flow (PFDF) prediction, uses an LLM to convert operational weather forecast narratives into structured meteorological indicators and augments features with terrain and soil-moisture context, and proposes a reinforcement-learning feature-selection framework that identifies key drivers by perturbing features until positive and negative events become indistinguishable; learned models consistently outperform the fixed-coefficient USGS logistic-regression baseline, augmented features further improve most model families, and storm variables dominate globally while feature importance varies substantially across regions.
University of Georgia University of Oregon
Zhisheng Qi（First author）
Post-Wildfire Debris Flow Prediction Reinforcement Learning Policy Gradient Methods
International Journal of Innovative Science and Research Technology (IJISRT) Sep 26, 2026
A review maps multi-agent reinforcement learning from Markov games to three algorithm lineages, flagging scale, unfamiliar-partner cooperation and deployment safety as open
This review uses Markov games and their decentralized, partially observed variants as the mathematical frame, groups the multi-agent reinforcement learning literature into three lineages—agents learning in isolation, methods that factor a team value into per-agent pieces (VDN, QMIX and their descendants), and actor-critic schemes trained against a critic with global knowledge (MADDPG, COMA, MAPPO)—discusses the obstacles that distinguish MARL from its single-agent counterpart, namely a moving-target learning problem, dividing a shared reward among team members, limited local views, and growth of the joint action space, explains why training with global information but acting on local information has become the standard design, reviews uses in real-time strategy games, robot teams, automate
This review uses Markov games and their decentralized, partially observed variants as the mathematical frame, groups the multi-agent reinforcement learning literature into three lineages—agents learning in isolation, methods that factor a team value into per-agent pieces (VDN, QMIX and their descendants), and actor-critic schemes trained against a critic with global knowledge (MADDPG, COMA, MAPPO)—discusses the obstacles that distinguish MARL from its single-agent counterpart, namely a moving-target learning problem, dividing a shared reward among team members, limited local views, and growth of the joint action space, explains why training with global information but acting on local information has become the standard design, reviews uses in real-time strategy games, robot teams, automate
Rajiv Ranganath（First author）
Multi-Agent Reinforcement Learning Value Decomposition Policy Gradient Methods Centralized Training with Decentralized Execution

You’re all caught up

Why curriculum learning pays off in reinforcement learning: sparse rewards can stall learning without one, while dense rewards gain little

Unified Benchmark Plus RL Feature Selection: Learned Models Consistently Beat Fixed-Coefficient USGS Model for Post-Wildfire Debris Flow Prediction, with Drivers Varying Sharply by Region

A review maps multi-agent reinforcement learning from Markov games to three algorithm lineages, flagging scale, unfamiliar-partner cooperation and deployment safety as open