Cover Image

Adaptive Layerwise Perturbation: Unifying Off-Policy Corrections for LLM RL

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ye, Chenlu, Zhang, Xuanchang, Hao, Yifan, Yu, Zhou, Zhang, Ziji, Gullapalli, Abhinav, Chen, Hao, Huang, Jing, Zhang, Tong
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

Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
by: Ye, Chenlu, et al.
Published: (2025)

Transformers as Multi-task Learners: Decoupling Features in Hidden Markov Models
by: Hao, Yifan, et al.
Published: (2025)

Understanding Overadaptation in Supervised Fine-Tuning: The Role of Ensemble Methods
by: Hao, Yifan, et al.
Published: (2025)

Stable Asynchrony: Variance-Controlled Off-Policy RL for LLMs
by: Huang, Luke J., et al.
Published: (2026)

Reinforce-Ada: An Adaptive Sampling Framework under Non-linear RL Objectives
by: Xiong, Wei, et al.
Published: (2025)

Offline-Boosted Actor-Critic: Adaptively Blending Optimal Historical Behaviors in Deep Off-Policy RL
by: Luo, Yu, et al.
Published: (2024)

Trust the Batch, On- or Off-Policy: Adaptive Policy Optimization for RL Post-Training
by: Fakoor, Rasool, et al.
Published: (2026)

Soft Policy Optimization: Online Off-Policy RL for Sequence Models
by: Cohen, Taco, et al.
Published: (2025)

Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective
by: Zhang, Yuheng, et al.
Published: (2026)

How and Where to Translate? The Impact of Translation Strategies in Cross-lingual LLM Prompting
by: Gupta, Aman, et al.
Published: (2025)

On-Policy RL Meets Off-Policy Experts: Harmonizing Supervised Fine-Tuning and Reinforcement Learning via Dynamic Weighting
by: Zhang, Wenhao, et al.
Published: (2025)

Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction
by: Guan, Zhong, et al.
Published: (2026)

OMPO: A Unified Framework for RL under Policy and Dynamics Shifts
by: Luo, Yu, et al.
Published: (2024)

Optimizing Chain-of-Thought Reasoners via Gradient Variance Minimization in Rejection Sampling and RL
by: Yao, Jiarui, et al.
Published: (2025)

Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning
by: Zhao, Shiwan, et al.
Published: (2026)

DLP: Dynamic Layerwise Pruning in Large Language Models
by: Chen, Yuli, et al.
Published: (2025)

Self-rewarding correction for mathematical reasoning
by: Xiong, Wei, et al.
Published: (2025)

USP-Gaussian: Unifying Spike-based Image Reconstruction, Pose Correction and Gaussian Splatting
by: Chen, Kang, et al.
Published: (2024)

ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents
by: Zhang, Hao, et al.
Published: (2026)

WoVR: World Models as Reliable Simulators for Post-Training VLA Policies with RL
by: Jiang, Zhennan, et al.
Published: (2026)

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

Your Reward Function for RL is Your Best PRM for Search: Unifying RL and Search-Based TTS
by: Jin, Can, et al.
Published: (2025)

BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping
by: Xi, Zhiheng, et al.
Published: (2025)

Taming Latency-Memory Trade-Off in MoE-Based LLM Serving via Fine-Grained Expert Offloading
by: Yu, Hanfei, et al.
Published: (2025)

Policy Learning for Off-Dynamics RL with Deficient Support
by: Van, Linh Le Pham, et al.
Published: (2024)

When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs
by: Zeng, Yifan, et al.
Published: (2026)

LLM-CAS: Dynamic Neuron Perturbation for Real-Time Hallucination Correction
by: Zhang, Jensen, et al.
Published: (2025)

ELABORATION: A Comprehensive Benchmark on Human-LLM Competitive Programming
by: Yang, Xinwei, et al.
Published: (2025)

Scaling up Multi-Turn Off-Policy RL and Multi-Agent Tree Search for LLM Step-Provers
by: Xin, Ran, et al.
Published: (2025)

Kernel Metric Learning for In-Sample Off-Policy Evaluation of Deterministic RL Policies
by: Lee, Haanvid, et al.
Published: (2024)

A Unifying View of Coverage in Linear Off-Policy Evaluation
by: Amortila, Philip, et al.
Published: (2026)

TrajLLM: A Modular LLM-Enhanced Agent-Based Framework for Realistic Human Trajectory Simulation
by: Ju, Chenlu, et al.
Published: (2025)

Character-Level Perturbations Disrupt LLM Watermarks
by: Zhang, Zhaoxi, et al.
Published: (2025)

LISA: Layerwise Importance Sampling for Memory-Efficient Large Language Model Fine-Tuning
by: Pan, Rui, et al.
Published: (2024)

Asynchronous RLHF: Faster and More Efficient Off-Policy RL for Language Models
by: Noukhovitch, Michael, et al.
Published: (2024)

VESPO: Variational Sequence-Level Soft Policy Optimization for Stable Off-Policy LLM Training
by: Shen, Guobin, et al.
Published: (2026)

SOPE: Stabilizing Off-Policy Evaluation for Online RL with Prior Data
by: Romeo, Carlo, et al.
Published: (2026)

LLMRouterBench: A Massive Benchmark and Unified Framework for LLM Routing
by: Li, Hao, et al.
Published: (2026)

Outlier-weighed Layerwise Sampling for LLM Fine-tuning
by: Li, Pengxiang, et al.
Published: (2024)

MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent
by: Yu, Hongli, et al.
Published: (2025)