English
Related papers

Related papers: Hail to the Thief: Exploring Attacks and Defenses …

200 papers

Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks. Direct preference optimization (DPO), a widely deployed alignment method, exhibits limitations in both experimental…

Computation and Language · Computer Science 2025-10-31 Xuandong Zhao , Will Cai , Tianneng Shi , David Huang , Licong Lin , Song Mei , Dawn Song

Group Relative Policy Optimization (GRPO) has recently emerged as a practical recipe for aligning large language models with verifiable objectives. However, under sparse terminal rewards, GRPO often stalls because rollouts within a group…

Machine Learning · Computer Science 2026-02-04 Baohao Liao , Hanze Dong , Xinxing Xu , Christof Monz , Jiang Bian

Group-based reinforcement learning methods, like Group Relative Policy Optimization (GRPO), are widely used nowadays to post-train large language models. Despite their empirical success, they exhibit structural mismatches between reward…

Machine Learning · Computer Science 2026-01-09 Aleksandar Fontana , Marco Simoni , Giulio Rossolini , Andrea Saracino , Paolo Mori

Diffusion large language models (dLLMs), which offer a promising alternative to traditional autoregressive LLMs, have recently shown strong results in pretraining. However, due to their lack of tractable sequence-level likelihoods, they…

Machine Learning · Computer Science 2026-02-03 Anthony Zhan

Group Relative Policy Optimization (GRPO) has emerged as a popular algorithm for reinforcement learning with large language models (LLMs). However, upon analyzing its clipping mechanism, we argue that it is suboptimal in certain scenarios.…

Machine Learning · Computer Science 2026-01-08 Chi Liu , Xin Chen

Large language models (LLMs) employ safety mechanisms to prevent harmful outputs, yet these defenses primarily rely on semantic pattern matching. We show that encoding harmful prompts as coherent mathematical problems -- using formalisms…

Cryptography and Security · Computer Science 2026-05-06 Haoyu Zhang , Mohammad Zandsalimy , Shanu Sushmita

Training Large Language Models (LLMs) for multi-turn Tool-Integrated Reasoning (TIR) - where models iteratively reason, generate code, and verify through execution - remains challenging for existing reinforcement learning (RL) approaches.…

Machine Learning · Computer Science 2026-04-21 Yifeng Ding , Hung Le , Songyang Han , Kangrui Ruan , Zhenghui Jin , Varun Kumar , Zijian Wang , Anoop Deoras

Group Relative Policy Optimization (GRPO) is widely used for reinforcement learning with verifiable rewards, but it often suffers from advantage collapse: when all rollouts in a group receive the same reward, the group yields zero relative…

Machine Learning · Computer Science 2026-04-02 Yu Xia , Canwen Xu , Zhewei Yao , Julian McAuley , Yuxiong He

Hybrid Group Relative Policy Optimization (Hybrid GRPO) is a reinforcement learning framework that extends Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO) by incorporating empirical multi-sample action…

Machine Learning · Computer Science 2025-02-05 Soham Sane

The Group Relative Policy Optimization (GRPO) algorithm has demonstrated considerable success in enhancing the reasoning capabilities of large language models (LLMs), as evidenced by DeepSeek-R1. However, the absence of intermediate…

Machine Learning · Computer Science 2025-06-06 Fei Ding , Baiqiao Wang , Zijian Zeng , Youwei Wang

Reinforcement learning (RL) has become a cornerstone for fine-tuning Large Language Models (LLMs), with Proximal Policy Optimization (PPO) serving as the de facto standard algorithm. Despite its ubiquity, we argue that the core ratio…

Machine Learning · Computer Science 2026-05-27 Penghui Qi , Xiangxin Zhou , Zichen Liu , Tianyu Pang , Chao Du , Min Lin , Wee Sun Lee

Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as a powerful paradigm for facilitating the self-improvement of large language models (LLMs), particularly in the domain of complex reasoning tasks. However,…

Machine Learning · Computer Science 2025-07-17 Ziru Liu , Cheng Gong , Xinyu Fu , Yaofang Liu , Ran Chen , Shoubo Hu , Suiyun Zhang , Rui Liu , Qingfu Zhang , Dandan Tu

In an era where tool-augmented AI agents are becoming increasingly vital, our findings highlight the ability of Group Relative Policy Optimization (GRPO) to empower SLMs, which are traditionally constrained in tool use. The ability to use…

Computation and Language · Computer Science 2025-09-10 Dhruvi Paprunia , Vansh Kharidia , Pankti Doshi

Reinforcement learning methods such as GRPO have seen great popularity in LLM post-training. In GRPO, models produce completions to a set of prompts, which are rewarded, and the policy is updated towards the relatively high reward…

Machine Learning · Computer Science 2026-05-22 Nikolay Blagoev , Oğuzhan Ersoy , Wendelin Boehmer , Lydia Yiyu Chen

Recent advancements in Reinforcement Learning with Human Feedback (RLHF) have significantly impacted the alignment of Large Language Models (LLMs). The sensitivity of reinforcement learning algorithms such as Proximal Policy Optimization…

Machine Learning · Computer Science 2025-06-10 Pankayaraj Pathmanathan , Souradip Chakraborty , Xiangyu Liu , Yongyuan Liang , Furong Huang

Reinforcement learning (RL) has proven effective in strengthening the reasoning capabilities of large language models (LLMs). A widely adopted method, Group Relative Policy Optimization (GRPO), has shown strong empirical results in training…

Machine Learning · Computer Science 2026-03-11 Peter Chen , Xiaopeng Li , Ziniu Li , Xi Chen , Tianyi Lin

Recent advances in large language models (LLMs) have broadened their applicability across diverse tasks, yet specialized domains still require targeted post training. Among existing methods, Group Relative Policy Optimization (GRPO) stands…

Machine Learning · Computer Science 2025-08-08 Ziyin Gu , Jingyao Wang , Ran Zuo , Chuxiong Sun , Zeen Song , Changwen Zheng , Wenwen Qiang

Group-Relative Policy Optimization (GRPO) is a key technique for training large reasoning models, yet it suffers from a critical vulnerability: the \emph{Think-Answer Mismatch}, where noisy reward signals corrupt the learning process. This…

Machine Learning · Computer Science 2025-08-11 Si Shen , Peijun Shen , Wenhua Zhao , Danhao Zhu

Recent advances in large language models have highlighted the critical need for precise control over model outputs through predefined constraints. While existing methods attempt to achieve this through either direct instruction-response…

Computation and Language · Computer Science 2025-03-27 Zhouhong Gu , Xingzhou Chen , Xiaoran Shi , Tao Wang , Suhang Zheng , Tianyu Li , Hongwei Feng , Yanghua Xiao

Large language models (LLMs) trained on webscale data can produce toxic outputs, raising concerns for safe deployment. Prior defenses, based on applications of DPO, NPO, and similar algorithms, reduce the likelihood of harmful…