Image de couverture de livre

What is the Alignment Objective of GRPO?

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Vojnovic, Milan, Yun, Se-Young
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

An Adaptive Approach for Infinitely Many-armed Bandits under Generalized Rotting Constraints
par: Kim, Jung-hun, et autres
Publié: (2024)

Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models
par: Park, Youngrok, et autres
Publié: (2025)

On the Convergence of Loss and Uncertainty-based Active Learning Algorithms
par: Haimovich, Daniel, et autres
Publié: (2023)

Improving Visual Representation Alignment Generation with GRPO
par: Mo, Shentong, et autres
Publié: (2026)

What Is the Alignment Tax?
par: Young, Robin
Publié: (2026)

AMIR-GRPO: Inducing Implicit Preference Signals into GRPO
par: Yari, Amir Hossein, et autres
Publié: (2026)

GL-LowPopArt: A Nearly Instance-Wise Minimax-Optimal Estimator for Generalized Low-Rank Trace Regression
par: Lee, Junghyun, et autres
Publié: (2025)

LVRPO: Language-Visual Alignment with GRPO for Multimodal Understanding and Generation
par: Mo, Shentong, et autres
Publié: (2026)

Predicting LLM Reasoning Performance with Small Proxy Model
par: Koh, Woosung, et autres
Publié: (2025)

IB-GRPO: Aligning LLM-based Learning Path Recommendation with Educational Objectives via Indicator-Based Group Relative Policy Optimization
par: Wang, Shuai, et autres
Publié: (2026)

DistiLLM: Towards Streamlined Distillation for Large Language Models
par: Ko, Jongwoo, et autres
Publié: (2024)

Revisiting Multi-Agent Debate as Test-Time Scaling: A Systematic Study of Conditional Effectiveness
par: Yang, Yongjin, et autres
Publié: (2025)

Automated Skill Discovery for Language Agents through Exploration and Iterative Feedback
par: Yang, Yongjin, et autres
Publié: (2025)

GRPO is Secretly a Process Reward Model
par: Sullivan, Michael, et autres
Publié: (2025)

Reward-free Alignment for Conflicting Objectives
par: Chen, Peter, et autres
Publié: (2026)

Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment
par: Hou, Xiaoyang, et autres
Publié: (2026)

Uncalibrated Reasoning: GRPO Induces Overconfidence for Stochastic Outcomes
par: Bereket, Michael, et autres
Publié: (2025)

GRPO-$λ$: Credit Assignment improves LLM Reasoning
par: Parthasarathi, Prasanna, et autres
Publié: (2025)

C-MORAL: Controllable Multi-Objective Molecular Optimization with Reinforcement Alignment for LLMs
par: Gao, Rui, et autres
Publié: (2026)

Meta-Aligner: Bidirectional Preference-Policy Optimization for Multi-Objective LLMs Alignment
par: Xu, Wenzhe, et autres
Publié: (2026)

Contextual Linear Bandits under Noisy Features: Towards Bayesian Oracles
par: Kim, Jung-hun, et autres
Publié: (2017)

mSFT: Addressing Dataset Mixtures Overfitting Heterogeneously in Multi-task SFT
par: Koh, Woosung, et autres
Publié: (2026)

Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients
par: Mansouri, Omar El, et autres
Publié: (2025)

S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models
par: Dai, Muzhi, et autres
Publié: (2025)

Pareto Multi-Objective Alignment for Language Models
par: He, Qiang, et autres
Publié: (2025)

A Unified Framework for Rethinking Policy Divergence Measures in GRPO
par: Wu, Qingyuan, et autres
Publié: (2026)

Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO
par: Ren, Yiming, et autres
Publié: (2026)

A Unified Pair-GRPO Family: From Implicit to Explicit Preference Constraints for Stable and General RL Alignment
par: Yu, Hao
Publié: (2026)

Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization
par: Zhou, Zhanhui, et autres
Publié: (2023)

Automated Filtering of Human Feedback Data for Aligning Text-to-Image Diffusion Models
par: Yang, Yongjin, et autres
Publié: (2024)

CoRPO: Adding a Correctness Bias to GRPO Improves Generalization
par: Garg, Anisha, et autres
Publié: (2025)

Prefix Grouper: Efficient GRPO Training through Shared-Prefix Forward
par: Liu, Zikang, et autres
Publié: (2025)

PARM: Multi-Objective Test-Time Alignment via Preference-Aware Autoregressive Reward Model
par: Lin, Baijiong, et autres
Publié: (2025)

Adaptive Alignment: Dynamic Preference Adjustments via Multi-Objective Reinforcement Learning for Pluralistic AI
par: Harland, Hadassah, et autres
Publié: (2024)

Conditional Synthesis of 3D Molecules with Time Correction Sampler
par: Jung, Hojung, et autres
Publié: (2024)

Comparative Analysis and Parametric Tuning of PPO, GRPO, and DAPO for LLM Reasoning Enhancement
par: Lian, Yongsheng
Publié: (2025)

Discovering Hidden Algebraic Structures via Transformers with Rank-Aware Beam GRPO
par: Lee, Jaeha, et autres
Publié: (2025)

GROW: Aligning GRPO with State-Action Modeling for Open-World VLM Agents
par: Wu, Xiongbin, et autres
Publié: (2026)

Unveiling Implicit Advantage Symmetry: Why GRPO Struggles with Exploration and Difficulty Adaptation
par: Yu, Zhiqi, et autres
Publié: (2026)

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare
par: Plyusov, Daniil, et autres
Publié: (2026)