相关论文: GRPO-Guard: Mitigating Implicit Over-Optimization …

Reinforcement learning has significantly enhanced the reasoning capabilities of Large Language Models (LLMs) in complex problem-solving tasks. Recently, the introduction of DeepSeek R1 has inspired a surge of interest in leveraging…

Despite Proximal Policy Optimization (PPO) dominating policy gradient methods -- from robotic control to game AI -- its static trust region forces a brittle trade-off: aggressive clipping stifles early exploration, while late-stage updates…

RLVR has enhanced the reasoning capabilities of Large Language Models (LLMs) across various tasks. However, GRPO, a representative RLVR algorithm, suffers from a critical limitation: when all responses within a group are either entirely…

Among on-policy reinforcement learning algorithms, Proximal Policy Optimization (PPO) demonstrates is widely favored for its simplicity, numerical stability, and strong empirical performance. Standard PPO relies on surrogate objectives…

Reinforcement learning, including reinforcement learning with verifiable rewards (RLVR), has emerged as a powerful approach for LLM post-training. Central to these approaches is the design of the importance sampling (IS) ratio used in…

Group Relative Policy Optimization has emerged as essential for aligning video diffusion models with human preferences, but faces a critical computational bottleneck: training a 14B parametered model typically demands hundreds of GPU days…

Group Relative Policy Optimization (GRPO) has emerged as an effective method for training reasoning models. While it computes advantages based on group mean, GRPO treats each output as an independent sample during the optimization and…

We study reinforcement learning in hybrid discrete-continuous action spaces, such as settings where the discrete component selects a regime (or index) and the continuous component optimizes within it -- a structure common in robotics,…

Process reward models (PRMs) allow for fine-grained credit assignment in reinforcement learning (RL), and seemingly contrast with outcome reward models (ORMs), which assign a single reward to an entire trajectory. However, we provide…

On-policy reinforcement learning methods, like Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO), often demand extensive data per update, leading to sample inefficiency. This paper introduces Reflective Policy…

Reinforcement Learning (RL) methods, exemplified by Group Relative Policy Optimization (GRPO) and its variants, play a central role in developing reasoning models. However, these methods often suffer from a critical overconfidence issue,…

Recent advances in generative AI have revolutionized visual content creation, yet aligning model outputs with human preferences remains a critical challenge. While Reinforcement Learning (RL) has emerged as a promising approach for…

Recent advances in flow matching models, particularly with reinforcement learning (RL), have significantly enhanced human preference alignment in few step text to image generators. However, existing RL based approaches for flow matching…

Policy optimization is a fundamental principle for designing reinforcement learning algorithms, and one example is the proximal policy optimization algorithm with a clipped surrogate objective (PPO-Clip), which has been popularly used in…

Existing Reinforcement Learning from Verifiable Rewards (RLVR) methods, such as Group Relative Policy Optimization (GRPO), have achieved remarkable progress in improving the reasoning capabilities of Large Reasoning Models (LRMs). However,…

The policy gradient method enjoys the simplicity of the objective where the agent optimizes the cumulative reward directly. Moreover, in the continuous action domain, parameterized distribution of action distribution allows easy control of…

Proximal Policy Optimization (PPO) is a widely used reinforcement learning algorithm that heavily relies on accurate advantage estimates for stable and efficient training. However, raw advantage signals can exhibit significant variance,…

In safety-critical domains, reinforcement learning (RL) agents must often satisfy strict, zero-cost safety constraints while accomplishing tasks. Existing model-free methods frequently either fail to achieve near-zero safety violations or…

Group Relative Policy Optimization (GRPO) is a powerful technique for aligning generative models, but its effectiveness is bottlenecked by the conflict between large group sizes and prohibitive computational costs. In this work, we…

Recent studies have demonstrated the efficacy of integrating Group Relative Policy Optimization (GRPO) into flow matching models, particularly for text-to-image and text-to-video generation. However, we find that directly applying these…