Imagen de Portada

SHAPE: Stage-aware Hierarchical Advantage via Potential Estimation for LLM Reasoning

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Ai, Zhengyang, Shan, Zikang, Ai, Xiaodong, Tang, Jingxian, Hu, Hangkai, Lu, Pinyan
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Arbitrage: Efficient Reasoning via Advantage-Aware Speculation
por: Maheswaran, Monishwaran, et al.
Publicado: (2025)

Revisiting Entropy Regularization: Adaptive Coefficient Unlocks Its Potential for LLM Reinforcement Learning
por: Zhang, Xiaoyun, et al.
Publicado: (2025)

Bringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning
por: Shan, Zikang, et al.
Publicado: (2026)

ReasonFlux: Hierarchical LLM Reasoning via Scaling Thought Templates
por: Yang, Ling, et al.
Publicado: (2025)

MathScale: Scaling Instruction Tuning for Mathematical Reasoning
por: Tang, Zhengyang, et al.
Publicado: (2024)

Less is More: Improving LLM Alignment via Preference Data Selection
por: Deng, Xun, et al.
Publicado: (2025)

Logical Structure as Knowledge: Enhancing LLM Reasoning via Structured Logical Knowledge Density Estimation
por: Bi, Zhen, et al.
Publicado: (2025)

GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation
por: Li, Sijia, et al.
Publicado: (2026)

Efficient Long-distance Latent Relation-aware Graph Neural Network for Multi-modal Emotion Recognition in Conversations
por: Shou, Yuntao, et al.
Publicado: (2024)

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

HS-STaR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation
por: Xiong, Feng, et al.
Publicado: (2025)

RuleReasoner: Reinforced Rule-based Reasoning via Domain-aware Dynamic Sampling
por: Liu, Yang, et al.
Publicado: (2025)

Blockwise Advantage Estimation for Multi-Objective RL with Verifiable Rewards
por: Pavlenko, Kirill, et al.
Publicado: (2026)

Unveiling Memorization-Generalization Coexistence: A Case Study on Arithmetic Tasks with Label Noise
por: Liu, Linyu, et al.
Publicado: (2026)

Graph Counselor: Adaptive Graph Exploration via Multi-Agent Synergy to Enhance LLM Reasoning
por: Gao, Junqi, et al.
Publicado: (2025)

No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shaping
por: Le, Thanh-Long V., et al.
Publicado: (2025)

Better LLM Reasoning via Dual-Play
por: Zhang, Zhengxin, et al.
Publicado: (2025)

How Numerical Precision Affects Arithmetical Reasoning Capabilities of LLMs
por: Feng, Guhao, et al.
Publicado: (2024)

Conditional Advantage Estimation for Reinforcement Learning in Large Reasoning Models
por: Chen, Guanxu, et al.
Publicado: (2025)

MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining
por: Wen, Bingbing, et al.
Publicado: (2026)

Tree-OPO: Off-policy Monte Carlo Tree-Guided Advantage Optimization for Multistep Reasoning
por: Huang, Bingning, et al.
Publicado: (2025)

Mitigating Reward Hacking in RLHF via Advantage Sign Robustness
por: Ono, Shinnosuke, et al.
Publicado: (2026)

RelayLLM: Efficient Reasoning via Collaborative Decoding
por: Huang, Chengsong, et al.
Publicado: (2026)

Advancing LLM Reasoning Generalists with Preference Trees
por: Yuan, Lifan, et al.
Publicado: (2024)

Option-ID Based Elimination For Multiple Choice Questions
por: Zhu, Zhenhao, et al.
Publicado: (2025)

Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM's Reasoning Capability
por: Lin, Zicheng, et al.
Publicado: (2024)

CoRT: Code-integrated Reasoning within Thinking
por: Li, Chengpeng, et al.
Publicado: (2025)

From Reasoning Chains to Verifiable Subproblems: Curriculum Reinforcement Learning Enables Credit Assignment for LLM Reasoning
por: Jiang, Xitai, et al.
Publicado: (2026)

Safe: Enhancing Mathematical Reasoning in Large Language Models via Retrospective Step-aware Formal Verification
por: Liu, Chengwu, et al.
Publicado: (2025)

Internalizing LLM Reasoning via Discovery and Replay of Latent Actions
por: Shi, Zhenning, et al.
Publicado: (2026)

Assessing LLM Reasoning Steps via Principal Knowledge Grounding
por: Hwang, Hyeon, et al.
Publicado: (2025)

GIFT: Reconciling Post-Training Objectives via Finite-Temperature Gibbs Initialization
por: Zhao, Zhengyang, et al.
Publicado: (2026)

FlowRL: Matching Reward Distributions for LLM Reasoning
por: Zhu, Xuekai, et al.
Publicado: (2025)

Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision
por: Xi, Zhiheng, et al.
Publicado: (2024)

CosmicFish-HRM: Adaptive Reasoning via Hierarchical Recurrent Mechanisms in Compact Language Models
por: Lakkapragada, Venkat Akhil
Publicado: (2026)

GraphDancer: Training LLMs to Explore and Reason over Graphs via Two-Stage Curriculum Post-Training
por: Bai, Yuyang, et al.
Publicado: (2026)

Stable Adaptive Thinking via Advantage Shaping and Length-Aware Gradient Regulation
por: Xu, Zihang, et al.
Publicado: (2026)

LiteStage: Latency-aware Layer Skipping for Multi-stage Reasoning
por: Kang, Beomseok, et al.
Publicado: (2025)

ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning
por: Potamitis, Nearchos, et al.
Publicado: (2025)

DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoning
por: Wang, Yaxuan, et al.
Publicado: (2025)