Cover Image

When to Stop Reusing: Dynamic Gradient Gating for Sample-Efficient RLVR

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Miao, Yuchun, Zhang, Sen, Zhang, Yuqi, Shi, Yaorui, Gu, Qi, Cai, Xunliang, Zhang, Lefei
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

InfoRM: Mitigating Reward Hacking in RLHF via Information-Theoretic Reward Modeling
by: Miao, Yuchun, et al.
Published: (2024)

The Energy Loss Phenomenon in RLHF: A New Perspective on Mitigating Reward Hacking
by: Miao, Yuchun, et al.
Published: (2025)

When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning
by: Sheng, Leheng, et al.
Published: (2026)

Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning
by: Shi, Yaorui, et al.
Published: (2026)

On the Implicit Reward Overfitting and the Low-rank Dynamics in RLVR
by: Ye, Hao, et al.
Published: (2026)

MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning
by: Fu, Xiaoliang, et al.
Published: (2026)

Rethinking the Sampling Criteria in Reinforcement Learning for LLM Reasoning: A Competence-Difficulty Alignment Perspective
by: Kong, Deyang, et al.
Published: (2025)

RLVR without Ineffective Samples: Group Prioritized Off-Policy Optimization for LLM Reasoning
by: Mao, Yixiu, et al.
Published: (2026)

Flexible Entropy Control in RLVR with a Gradient-Preserving Perspective
by: Chen, Kun, et al.
Published: (2026)

BEACON: Bayesian Optimal Stopping for Efficient LLM Sampling
by: Wan, Guangya, et al.
Published: (2025)

GeoRA: Geometry-Aware Low-Rank Adaptation for RLVR
by: Zhang, Jiaying, et al.
Published: (2026)

Asymmetric Advantage Modulation Calibrates Entropy Dynamics in RLVR
by: Gu, Hengrui, et al.
Published: (2026)

VL Norm: Rethink Loss Aggregation in RLVR
by: He, Zhiyuan, et al.
Published: (2025)

Lotus: Efficient LLM Training by Randomized Low-Rank Gradient Projection with Adaptive Subspace Switching
by: Miao, Tianhao, et al.
Published: (2026)

JudgeRLVR: Judge First, Generate Second for Efficient Reasoning
by: Duo, Jiangshan, et al.
Published: (2026)

Reasoner for Real-World Event Detection: Scaling Reinforcement Learning via Adaptive Perplexity-Aware Sampling Strategy
by: Zhang, Xiaoyun, et al.
Published: (2025)

Beyond Uniform Credit Assignment: Selective Eligibility Traces for RLVR
by: Mou, Chaoli, et al.
Published: (2026)

Quantile Advantage Estimation: Stabilizing RLVR for LLM Reasoning
by: Wu, Junkang, et al.
Published: (2025)

Open-Medical-R1: How to Choose Data for RLVR Training at Medicine Domain
by: Qiu, Zhongxi, et al.
Published: (2025)

Mitigating Tail Narrowing in LLM Self-Improvement via Socratic-Guided Sampling
by: Ding, Yiwen, et al.
Published: (2024)

RL in the Wild: Characterizing RLVR Training in LLM Deployment
by: Zhou, Jiecheng, et al.
Published: (2025)

Beyond Importance Sampling: Rejection-Gated Policy Optimization
by: Sun, Ziwu, et al.
Published: (2026)

Shorter but not Worse: Frugal Reasoning via Easy Samples as Length Regularizers in Math RLVR
by: Bounhar, Abdelaziz, et al.
Published: (2025)

IRDS: Interpretable RLVR Data Selection via Verifier-Coupled Sparse Autoencoder Coverage
by: Li, Yuhan, et al.
Published: (2026)

Information-Theoretic Reward Modeling for Stable RLHF: Detecting and Mitigating Reward Hacking
by: Miao, Yuchun, et al.
Published: (2025)

COSEE: Consistency-Oriented Signal-Based Early Exiting via Calibrated Sample Weighting Mechanism
by: He, Jianing, et al.
Published: (2024)

The Path Not Taken: RLVR Provably Learns Off the Principals
by: Zhu, Hanqing, et al.
Published: (2025)

Language-Enhanced Representation Learning for Single-Cell Transcriptomics
by: Shi, Yaorui, et al.
Published: (2025)

Adaptive Negative Reinforcement for LLM Reasoning:Dynamically Balancing Correction and Diversity in RLVR
by: Ingle, Yash, et al.
Published: (2026)

Generalized Policy Improvement Algorithms with Theoretically Supported Sample Reuse
by: Queeney, James, et al.
Published: (2022)

Reasoning-Enhanced Large Language Models for Molecular Property Prediction
by: Zhuang, Jiaxi, et al.
Published: (2025)

EntropyStop: Unsupervised Deep Outlier Detection with Loss Entropy
by: Huang, Yihong, et al.
Published: (2024)

Listwise Policy Optimization: Group-based RLVR as Target-Projection on the LLM Response Simplex
by: Qu, Yun, et al.
Published: (2026)

Gradient-Informed Temporal Sampling Improves Rollout Accuracy in PDE Surrogate Training
by: Wang, Wenshuo, et al.
Published: (2026)

CoBA-RL: Capability-Oriented Budget Allocation for Reinforcement Learning in LLMs
by: Yao, Zhiyuan, et al.
Published: (2026)

CADENT: Gated Hybrid Distillation for Sample-Efficient Transfer in Reinforcement Learning
by: Alinejad, Mahyar, et al.
Published: (2026)

Beyond Variance: Prompt-Efficient RLVR via Rare-Event Amplification and Bidirectional Pairing
by: Pang, Yujuan, et al.
Published: (2026)

$V_{0.5}$: Generalist Value Model as a Prior for Sparse RL Rollouts
by: Zhang, Yi-Kai, et al.
Published: (2026)

When Does Learning Renormalize? Sufficient Conditions for Power Law Spectral Dynamics
by: Zhang, Yizhou
Published: (2025)

Matrix-Space Reinforcement Learning for Reusing Local Transition Geometry
by: Zhang, Zuyuan, et al.
Published: (2026)