Imagen de Portada

Why GRPO Needs Normalization: A Local-Curvature Perspective on Adaptive Gradients

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Ge, Cheng, Yin, Caitlyn Heqi, Liang, Hao, Zhang, Jiawei
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Gradient Starvation in Binary-Reward GRPO: Why Group-Mean Centering Fails and Why the Simplest Fix Works
por: Nie, Wenhua, et al.
Publicado: (2026)

DRA-GRPO: Your GRPO Needs to Know Diverse Reasoning Paths for Mathematical Reasoning
por: Chen, Xiwen, et al.
Publicado: (2025)

Is Conformal Factuality for RAG-based LLMs Robust? Novel Metrics and Systematic Insights
por: Chen, Yi, et al.
Publicado: (2026)

Why Transformers Need Adam: A Hessian Perspective
por: Zhang, Yushun, et al.
Publicado: (2024)

Unveiling Implicit Advantage Symmetry: Why GRPO Struggles with Exploration and Difficulty Adaptation
por: Yu, Zhiqi, et al.
Publicado: (2026)

Why Do We Need Warm-up? A Theoretical Perspective
por: Alimisis, Foivos, et al.
Publicado: (2025)

Adaptive Optimization via Momentum on Variance-Normalized Gradients
por: Patitucci, Francisco, et al.
Publicado: (2026)

Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients
por: Mansouri, Omar El, et al.
Publicado: (2025)

Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients
por: Zhang, Zhenyu, et al.
Publicado: (2024)

LithoGRPO: Fast Inverse Lithography via GRPO Reinforced Flow Matching
por: Lai, Yao, et al.
Publicado: (2026)

A Locally Adaptive Algorithm for Multiple Testing with Network Structure
por: Liang, Ziyi, et al.
Publicado: (2022)

Stochastic Normalized Gradient Descent with Momentum for Large-Batch Training
por: Zhao, Shen-Yi, et al.
Publicado: (2020)

Balanced Aggregation: Understanding and Fixing Aggregation Bias in GRPO
por: Zeng, Zhiyuan, et al.
Publicado: (2026)

Minimax Optimality of Score-based Diffusion Models: Beyond the Density Lower Bound Assumptions
por: Zhang, Kaihong, et al.
Publicado: (2024)

Reconstructing MODIS Normalized Difference Snow Index Product on Greenland Ice Sheet Using Spatiotemporal Extreme Gradient Boosting Model
por: Ye, Fan, et al.
Publicado: (2024)

Adaptive Gradient Normalization and Independent Sampling for (Stochastic) Generalized-Smooth Optimization
por: Yang, Yufeng, et al.
Publicado: (2024)

AMIR-GRPO: Inducing Implicit Preference Signals into GRPO
por: Yari, Amir Hossein, et al.
Publicado: (2026)

GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning
por: Chen, Yi, et al.
Publicado: (2025)

Normal Guidance is what Attention Needs
por: Harvey, Ethan, et al.
Publicado: (2026)

Mastering AI: Big Data, Deep Learning, and the Evolution of Large Language Models -- AutoML from Basics to State-of-the-Art Techniques
por: Feng, Pohsun, et al.
Publicado: (2024)

FairGRPO: Fair Reinforcement Learning for Equitable Clinical Reasoning
por: Dai, Shiqi, et al.
Publicado: (2025)

AceGRPO: Adaptive Curriculum Enhanced Group Relative Policy Optimization for Autonomous Machine Learning Engineering
por: Cai, Yuzhu, et al.
Publicado: (2026)

How Does Gradient Descent Learn Features -- A Local Analysis for Regularized Two-Layer Neural Networks
por: Zhou, Mo, et al.
Publicado: (2024)

The Blessing of Dimensionality in LLM Fine-tuning: A Variance-Curvature Perspective
por: Liang, Qiyao, et al.
Publicado: (2026)

Hard Examples Are All You Need: Maximizing GRPO Post-Training Under Annotation Budgets
por: Pikus, Benjamin, et al.
Publicado: (2025)

When and Why Does Unsupervised RL Succeed in Mathematical Reasoning? A Manifold Envelopment Perspective
por: Zhang, Zelin, et al.
Publicado: (2026)

Unsupervised Adaptive Normalization
por: Faye, Bilal, et al.
Publicado: (2024)

Why Adam Can Beat SGD: Second-Moment Normalization Yields Sharper Tails
por: Jin, Ruinan, et al.
Publicado: (2026)

Advances in GRPO for Generation Models: A Survey
por: Liu, Zexiang, et al.
Publicado: (2026)

MAC: An Efficient Gradient Preconditioning using Mean Activation Approximated Curvature
por: Seung, Hyunseok, et al.
Publicado: (2025)

CAGE: Curvature-Aware Gradient Estimation For Accurate Quantization-Aware Training
por: Tabesh, Soroush, et al.
Publicado: (2025)

A Lightweight and Gradient-Stable Neural Layer
por: Yu, Yueyao, et al.
Publicado: (2021)

F-GRPO: Factorized Group-Relative Policy Optimization for Unified Candidate Generation and Ranking
por: Surana, Rohan, et al.
Publicado: (2026)

Exploiting Curvature in Online Convex Optimization with Delayed Feedback
por: Qiu, Hao, et al.
Publicado: (2025)

GRPO-RM: Fine-Tuning Representation Models via GRPO-Driven Reinforcement Learning
por: Xu, Yanchen, et al.
Publicado: (2025)

Position: Why a Dynamical Systems Perspective is Needed to Advance Time Series Modeling
por: Durstewitz, Daniel, et al.
Publicado: (2026)

CAT: Curvature-Adaptive Transformers for Geometry-Aware Learning
por: Lin, Ryan Y., et al.
Publicado: (2025)

Transformers Efficiently Perform In-Context Logistic Regression via Normalized Gradient Descent
por: Zhang, Chenyang, et al.
Publicado: (2026)

How Off-Policy Can GRPO Be? Mu-GRPO for Efficient LLM Reinforcement Learning
por: Tian, Minghao, et al.
Publicado: (2026)

Graph-GRPO: Training Graph Flow Models with Reinforcement Learning
por: Zhu, Baoheng, et al.
Publicado: (2026)