Cover Image

Reshaping Reasoning in LLMs: A Theoretical Analysis of RL Training Dynamics through Pattern Selection

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Chen, Xingwu, Li, Tianle, Zou, Difan
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Towards Theoretical Understanding of Transformer Test-Time Computing: Investigation on In-Context Linear Regression
by: Chen, Xingwu, et al.
Published: (2025)

On the Robustness of Transformers against Context Hijacking for Linear Classification
by: Li, Tianle, et al.
Published: (2025)

Structured Role-Aware Policy Optimization for Multimodal Reasoning
by: Jiang, Bingqing, et al.
Published: (2026)

A Convergence Analysis of Adaptive Optimizers under Floating-point Quantization
by: Tang, Xuan, et al.
Published: (2025)

A Human-Like Reasoning Framework for Multi-Phases Planning Task with Large Language Models
by: Xie, Chengxing, et al.
Published: (2024)

SortedRL: Accelerating RL Training for LLMs through Online Length-Aware Scheduling
by: Zhang, Yiqi, et al.
Published: (2026)

Train at Moving Edge: Online-Verified Prompt Selection for Efficient RL Training of Large Reasoning Model
by: Wu, Jiahao, et al.
Published: (2026)

On the Memorization of Consistency Distillation for Diffusion Models
by: Jiang, Bingqing, et al.
Published: (2026)

What Can Transformer Learn with Varying Depth? Case Studies on Sequence Learning Tasks
by: Chen, Xingwu, et al.
Published: (2024)

RL-STaR: Theoretical Analysis of Reinforcement Learning Frameworks for Self-Taught Reasoner
by: Chang, Fu-Chieh, et al.
Published: (2024)

Dynamics-Predictive Sampling for Active RL Finetuning of Large Reasoning Models
by: Mao, Yixiu, et al.
Published: (2026)

ReST-RL: Achieving Accurate Code Reasoning of LLMs with Optimized Self-Training and Decoding
by: Zhoubian, Sining, et al.
Published: (2025)

Dynamic Vocabulary Pruning: Stable LLM-RL by Taming the Tail
by: Li, Yingru, et al.
Published: (2025)

A Comedy of Estimators: On KL Regularization in RL Training of LLMs
by: Shah, Vedant, et al.
Published: (2025)

Towards Understanding Fine-Tuning Mechanisms of LLMs via Circuit Analysis
by: Wang, Xu, et al.
Published: (2025)

Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training
by: Liu, Mingjie, et al.
Published: (2025)

Learning under Quantization for High-Dimensional Linear Regression
by: Zhang, Dechen, et al.
Published: (2025)

Imbalanced Gradients in RL Post-Training of Multi-Task LLMs
by: Wu, Runzhe, et al.
Published: (2025)

Small Generalizable Prompt Predictive Models Can Steer Efficient RL Post-Training of Large Reasoning Models
by: Qu, Yun, et al.
Published: (2026)

How Post-Training Reshapes LLMs: A Mechanistic View on Knowledge, Truthfulness, Refusal, and Confidence
by: Du, Hongzhe, et al.
Published: (2025)

Balancing Multimodal Learning through Label Space Reshaping
by: Ma, Xiaoyu, et al.
Published: (2026)

Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders
by: Wang, Xu, et al.
Published: (2025)

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

Understanding the Dynamics of Demonstration Conflict in In-Context Learning
by: Jiao, Difan, et al.
Published: (2026)

The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs
by: Chen, Zihan, et al.
Published: (2026)

How Transformers Utilize Multi-Head Attention in In-Context Learning? A Case Study on Sparse Linear Regression
by: Chen, Xingwu, et al.
Published: (2024)

Beyond Alignment: Expanding Reasoning Capacity via Manifold-Reshaping Policy Optimization
by: Wang, Dayu, et al.
Published: (2026)

Tool Zero: Training Tool-Augmented LLMs via Pure RL from Scratch
by: Zeng, Yirong, et al.
Published: (2025)

GAC: Stabilizing Asynchronous RL Training for LLMs via Gradient Alignment Control
by: Xu, Haofeng, et al.
Published: (2026)

Understanding the Generalization of Stochastic Gradient Adam in Learning Neural Networks
by: Tang, Xuan, et al.
Published: (2025)

On the Collapse Errors Induced by the Deterministic Sampler for Diffusion Models
by: Zhang, Yi, et al.
Published: (2025)

Scaling Laws for Precision in High-Dimensional Linear Regression
by: Zhang, Dechen, et al.
Published: (2026)

Fast and Accurate Probing of In-Training LLMs' Downstream Performances
by: Liu, Zhichen, et al.
Published: (2026)

Learning-Zone Energy: Online Data Selection for Efficient RL Post-Training
by: Cui, Peng, et al.
Published: (2026)

On the Optimal Reasoning Length for RL-Trained Language Models
by: Nohara, Daisuke, et al.
Published: (2026)

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

ADORA: Training Reasoning Models with Dynamic Advantage Estimation on Reinforcement Learning
by: Ren, Qingnan, et al.
Published: (2026)

The Two-Stage Decision-Sampling Hypothesis: Understanding the Emergence of Self-Reflection in RL-Trained LLMs
by: Zhao, Zibo, et al.
Published: (2026)

A Systematic Investigation of The RL-Jailbreaker in LLMs
by: Mohammedalamen, Montaser, et al.
Published: (2026)

QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch
by: Gu, Hao, et al.
Published: (2026)