English
Related papers

Related papers: Advantage Collapse in Group Relative Policy Optimi…

200 papers

We introduce Adversarial Policy Optimization (AdvPO), a novel solution to the pervasive issue of reward over-optimization in Reinforcement Learning from Human Feedback (RLHF) for Large Language Models (LLMs). Over-optimization occurs when a…

Machine Learning · Computer Science 2024-07-10 Xiaoying Zhang , Jean-Francois Ton , Wei Shen , Hongning Wang , Yang Liu

Recent large reasoning models (LRMs) driven by reinforcement learning algorithms (e.g., GRPO) have achieved remarkable performance on challenging reasoning tasks. However, these models suffer from overthinking, generating unnecessarily long…

Artificial Intelligence · Computer Science 2026-03-03 Gang Li , Yan Chen , Ming Lin , Tianbao Yang

Reinforcement Learning (RL) for constrained MDPs (CMDPs) is an increasingly important problem for various applications. Often, the average criterion is more suitable than the discounted criterion. Yet, RL for average-CMDPs (ACMDPs) remains…

Machine Learning · Computer Science 2024-05-27 Akhil Agnihotri , Rahul Jain , Haipeng Luo

Large reasoning models (LRMs) are typically trained using reinforcement learning with verifiable reward (RLVR) to enhance their reasoning abilities. In this paradigm, policies are updated using both positive and negative self-generated…

Computation and Language · Computer Science 2025-12-29 Xinyu Tang , Yuliang Zhan , Zhixun Li , Wayne Xin Zhao , Zhenduo Zhang , Zujie Wen , Zhiqiang Zhang , Jun Zhou

We revisit Group Relative Policy Optimization (GRPO) in both on-policy and off-policy optimization regimes. Our motivation comes from recent work on off-policy Proximal Policy Optimization (PPO), which improves training stability, sampling…

Enhancing the reasoning capabilities of large language models effectively using reinforcement learning (RL) remains a crucial challenge. Existing approaches primarily adopt two contrasting advantage estimation granularities: token-level…

Machine Learning · Computer Science 2025-10-22 Yiran Guo , Lijie Xu , Jie Liu , Dan Ye , Shuang Qiu

Image captioning tasks usually use two-stage training to complete model optimization. The first stage uses cross-entropy as the loss function for optimization, and the second stage uses self-critical sequence training (SCST) for…

Computer Vision and Pattern Recognition · Computer Science 2025-03-04 Xu Liang

In this note, we examine the aggregation of preferences achieved by the Group Policy Optimisation (GRPO) algorithm, a reinforcement learning method used to train advanced artificial intelligence models such as DeepSeek-R1-Zero and…

Machine Learning · Computer Science 2025-03-14 Milan Vojnovic , Se-Young Yun

Self-evaluation, a model's ability to assess the correctness of its own output, is crucial for Large Multimodal Models (LMMs) to achieve self-improvement in multi-turn conversations, yet largely absent in foundation models. Recent work has…

Machine Learning · Computer Science 2025-08-14 Wenkai Wang , Hongcan Guo , Zheqi Lv , Shengyu Zhang

We introduce Self-correction Relative Policy Optimization (ScRPO), a novel reinforcement learning framework designed to empower large language models with advanced mathematical reasoning capabilities through iterative self-reflection and…

Artificial Intelligence · Computer Science 2026-01-06 Lianrui Li , Dakuan Lu , Jiawei Shao , Xuelong Li

This thesis develops theoretical frameworks and algorithms that advance constrained reinforcement learning (RL) across control, preference learning, and alignment of large language models. The first contribution addresses constrained Markov…

Machine Learning · Computer Science 2025-12-12 Akhil Agnihotri

Medical artificial intelligence systems have achieved remarkable diagnostic capabilities, yet they consistently exhibit performance disparities across demographic groups, causing real-world harm to underrepresented populations. While recent…

Machine Learning · Computer Science 2025-10-24 Shiqi Dai , Wei Dai , Jiaee Cheong , Paul Pu Liang

Search agents achieve strong question-answering performance through multi-turn interactions with search engines, with Group Relative Policy Optimization (GRPO) being a widely used training algorithm. However, GRPO-style algorithms still…

Computation and Language · Computer Science 2026-05-28 Jiayi Wu , Ruobing Xie , Zeqian Huang , Lei Jiang , Can Xu , Kangyang Luo , Bochen Lin , Ming Gao , Xiang Li

Group Relative Policy Optimization (GRPO) is a standard algorithm for reinforcement learning from verifiable rewards, but its group-mean-centered advantage can fail under binary rewards. The failure mode is gradient starvation: when every…

Machine Learning · Computer Science 2026-05-11 Wenhua Nie , Jianan Wu , Junlin Liu , Ziwei Li , Zheng Lin , Zhang Zijian , Yilong Fan , Haoran Zheng , Jyh-Shing Roger Jang

Aligning large language models with pointwise absolute rewards has so far required online, on-policy algorithms such as PPO and GRPO. In contrast, simpler methods that can leverage offline or off-policy data, such as DPO and REBEL, are…

Machine Learning · Computer Science 2025-12-02 Simon Matrenok , Skander Moalla , Caglar Gulcehre

Reinforcement Learning with Verifiable Rewards (RLVR) has become the standard paradigm for LLM mathematical reasoning, with Group Relative Policy Optimization (GRPO) serving as the dominant algorithm. We identify two overlooked…

Machine Learning · Computer Science 2026-05-13 Mingxiong Lin , Zhangquan Gong , Maowen Tang , Qian Li , Chuangchuang Wang , Jian Ma , Sutian Huang , Kai Tang , Haonan Lu

Group Relative Policy Optimization (GRPO) has proven to be an effective tool for post-training language models (LMs). However, AI systems are increasingly expressed as modular programs that mix together multiple LM calls with distinct…

Large-scale reinforcement learning with verifiable rewards (RLVR) has demonstrated its effectiveness in harnessing the potential of large language models (LLMs) for single-turn reasoning tasks. In realistic reasoning scenarios, LLMs can…

Group relative policy optimization (GRPO), a core methodological component of DeepSeekMath and DeepSeek-R1, has emerged as a cornerstone for scaling reasoning capabilities of large language models. Despite its widespread adoption and the…

Machine Learning · Computer Science 2026-03-24 Hongyi Zhou , Kai Ye , Erhan Xu , Jin Zhu , Ying Yang , Shijin Gong , Chengchun Shi

On-policy reinforcement learning (RL), particularly Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), has become the dominant paradigm for fine-tuning large language models (LLMs). While policy ratio clipping…

Machine Learning · Computer Science 2026-01-08 Yu Luo , Shuo Han , Yihan Hu , Dong Li , Jianye Hao