English
Related papers

Related papers: Advantage Collapse in Group Relative Policy Optimi…

200 papers

We investigate a narrow but common failure mode of GRPO-style reinforcement learning in the context of sparse verifiable rewards: early updates contain more responses with negative advantages than those with positive advantages, while…

Machine Learning · Computer Science 2026-05-29 Mohamed Sana , Nicola Piovesan , Antonio De Domenico , Fadhel Ayed , Haozhe Zhang

Reinforcement Learning with Verifiable Rewards (RLVR) enhances Large Language Model (LLM) reasoning but suffers from advantage collapse on ``hard samples'' where all rollouts fail. This lack of variance eliminates crucial learning signals.…

Machine Learning · Computer Science 2026-05-08 Xinyu Lu , Kaiqi Zhang , Jinglin Yang , Boxi Cao , Yaojie Lu , Hongyu Lin , Min He , Xianpei Han , Le Sun

Group-based reinforcement learning (RL) methods have achieved remarkable success in improving the performance of large language models (LLMs) and have been rapidly extended to agentic tasks. However, their credit assignment relies heavily…

Machine Learning · Computer Science 2026-05-27 Xin Cheng , Shuo He , Lang Feng , HaiYang Xu , Ming Yan , Lei Feng , Bo An

Reinforcement learning from verifiable rewards (RLVR), especially with Group Relative Policy Optimization (GRPO), has shown strong potential for improving the reasoning capabilities of large vision-language models (LVLMs). However, in…

Artificial Intelligence · Computer Science 2026-05-11 Bingqing Jiang , Difan Zou

Reinforcement Learning with Verifiable Rewards (RLVR) has become a key approach for improving the reasoning abilities of large language models. However, widely used critic-free algorithms such as Group Relative Policy Optimization (GRPO)…

Machine Learning · Computer Science 2026-05-08 Chaoli Mou , Zhan Zhuang , Xinning Chen , Yu Zhang

Despite the efficacy of Direct Preference Optimization (DPO) in aligning Large Language Models (LLMs), reward hacking remains a pivotal challenge. This issue emerges when LLMs excessively reduce the probability of rejected completions to…

Computation and Language · Computer Science 2025-08-26 Chenxu Yang , Ruipeng Jia , Mingyu Zheng , Naibin Gu , Zheng Lin , Siyuan Chen , Weichong Yin , Hua Wu , Weiping Wang

Group Relative Policy Optimization (GRPO) enables stable and preference-oriented updates via group-wise comparisons for post-training video generation. However, GRPO directly optimizes reward-induced advantages. Under sustained…

Computer Vision and Pattern Recognition · Computer Science 2026-03-18 Rui Li , Yuanzhi Liang , Ziqi Ni , Haibing Huang , Chi Zhang , Xuelong Li

This paper introduces Completion Pruning Policy Optimization (CPPO) to accelerate the training of reasoning models based on Group Relative Policy Optimization (GRPO). GRPO, while effective, incurs high training costs due to the need to…

Artificial Intelligence · Computer Science 2025-11-11 Zhihang Lin , Mingbao Lin , Yuan Xie , Rongrong Ji

Traditional on-policy Reinforcement Learning with Verifiable Rewards (RLVR) frameworks suffer from experience waste and reward homogeneity, which directly hinders learning efficiency on difficult samples during large language models…

Artificial Intelligence · Computer Science 2026-03-17 Xu Wan , Yansheng Wang , Wenqi Huang , Mingyang Sun

Existing Reinforcement Learning with Verifiable Rewards (RLVR) algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanisms that are fundamentally misaligned with the complex optimization dynamics of Large…

Machine Learning · Computer Science 2026-04-21 Xiaoliang Fu , Jiaye Lin , Yangyi Fang , Binbin Zheng , Chaowen Hu , Zekai Shao , Cong Qin , Lu Pan , Ke Zeng , Xunliang Cai

Constrained reinforcement learning has achieved promising progress in safety-critical fields where both rewards and constraints are considered. However, constrained reinforcement learning methods face challenges in striking the right…

Machine Learning · Computer Science 2024-10-29 Jianmina Ma , Jingtian Ji , Yue Gao

Group Relative Policy Optimization (GRPO) has recently shown strong performance in post-training large language models and vision-language models. It raises a question of whether the GRPO also significantly promotes the test-time adaptation…

Computer Vision and Pattern Recognition · Computer Science 2026-05-06 Yujun Li , Hongyuan Zhang , Yuan Yuan

Group Relative Policy Optimization (GRPO) significantly enhances the reasoning performance of Large Language Models (LLMs). However, this success heavily relies on expensive external verifiers or human rules. Such dependency not only leads…

Computation and Language · Computer Science 2026-03-03 Nonghai Zhang , Weitao Ma , Zhanyu Ma , Jun Xu , Jiuchong Gao , Jinghua Hao , Renqing He , Jingwen Xu

Group Relative Policy Optimization (GRPO) is widely used for critic-free Large Language Model (LLM) post-training, but its KL regularization is usually implemented as a local loss-side token penalty. We show that this misses the…

Machine Learning · Computer Science 2026-05-26 Jiarui Yao , Ruida Wang , Hao Bai , Tong Zhang

This paper introduces Group Sequence Policy Optimization (GSPO), our stable, efficient, and performant reinforcement learning algorithm for training large language models. Unlike previous algorithms that adopt token-level importance ratios,…

Machine Learning · Computer Science 2025-07-29 Chujie Zheng , Shixuan Liu , Mingze Li , Xiong-Hui Chen , Bowen Yu , Chang Gao , Kai Dang , Yuqiong Liu , Rui Men , An Yang , Jingren Zhou , Junyang Lin

Group Relative Policy Optimization (GRPO) has significantly enhanced the reasoning capability of large language models by optimizing the arithmetic mean of token-level rewards. Unfortunately, GRPO is observed to suffer from unstable policy…

Computation and Language · Computer Science 2025-10-21 Yuzhong Zhao , Yue Liu , Junpeng Liu , Jingye Chen , Xun Wu , Yaru Hao , Tengchao Lv , Shaohan Huang , Lei Cui , Qixiang Ye , Fang Wan , Furu Wei

Balancing exploration and exploitation during reinforcement learning fine-tuning of generative models presents a critical challenge, as existing approaches rely on fixed divergence regularization that creates an inherent dilemma: strong…

Machine Learning · Computer Science 2025-10-22 Jiajun Fan , Tong Wei , Chaoran Cheng , Yuxin Chen , Ge Liu

We present Anchored Direct Preference Optimization (ADPO), a policy alignment method derived from first principles of KL-regularized reinforcement learning. Unlike standard approaches that treat the reference policy merely as a regularizer,…

Machine Learning · Computer Science 2026-01-13 Wang Zixian

The Group Relative Policy Optimization (GRPO), a reinforcement learning method used to fine-tune large language models (LLMs), has proved its effectiveness in practical applications such as DeepSeek-R1. It raises a question whether GRPO can…

Machine Learning · Computer Science 2025-11-20 Yanchen Xu , Ziheng Jiao , Hongyuan Zhang , Xuelong Li

Group Relative Policy Optimization (GRPO) has significantly advanced the reasoning ability of large language models (LLMs), particularly by boosting their mathematical performance. However, GRPO and related entropy-regularization methods…

Computation and Language · Computer Science 2025-10-13 Xingyu Lin , Yilin Wen , En Wang , Du Su , Wenbin Liu , Chenfu Bao , Zhonghou Lv