Image de couverture de livre

RL Grokking Recipe: How Does RL Unlock and Transfer New Algorithms in LLMs?

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Sun, Yiyou, Cao, Yuhan, Huang, Pohao, Bai, Haoyue, Hajishirzi, Hannaneh, Dziri, Nouha, Song, Dawn
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative Generalization
par: Sun, Yiyou, et autres
Publié: (2025)

Climbing the Ladder of Reasoning: What LLMs Can-and Still Can't-Solve after SFT?
par: Sun, Yiyou, et autres
Publié: (2025)

TurnWise: The Gap between Single- and Multi-turn Language Model Capabilities
par: Graf, Victoria, et autres
Publié: (2026)

Why and How LLMs Hallucinate: Connecting the Dots with Subsequence Associations
par: Sun, Yiyou, et autres
Publié: (2025)

APT: Adaptive Pruning and Tuning Pretrained Language Models for Efficient Training and Inference
par: Zhao, Bowen, et autres
Publié: (2024)

How RL Unlocks the Aha Moment in Geometric Interleaved Reasoning
par: Zhang, Xiangxiang, et autres
Publié: (2026)

MIRAGE-Bench: LLM Agent is Hallucinating and Where to Find Them
par: Zhang, Weichen, et autres
Publié: (2025)

Knapsack RL: Unlocking Exploration of LLMs via Optimizing Budget Allocation
par: Li, Ziniu, et autres
Publié: (2025)

JustRL: Scaling a 1.5B LLM with a Simple RL Recipe
par: He, Bingxiang, et autres
Publié: (2025)

BTR: Binary Token Representations for Efficient Retrieval Augmented Language Models
par: Cao, Qingqing, et autres
Publié: (2023)

Surfacing Semantic Orthogonality Across Model Safety Benchmarks: A Multi-Dimensional Analysis
par: Bennion, Jonathan, et autres
Publié: (2025)

Vero: An Open RL Recipe for General Visual Reasoning
par: Sarch, Gabriel, et autres
Publié: (2026)

Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questions
par: Wiegreffe, Sarah, et autres
Publié: (2024)

How and Why LLMs Generalize: A Fine-Grained Analysis of LLM Reasoning from Cognitive Behaviors to Low-Level Patterns
par: Bai, Haoyue, et autres
Publié: (2025)

The Synergy of LLMs & RL Unlocks Offline Learning of Generalizable Language-Conditioned Policies with Low-fidelity Data
par: Pouplin, Thomas, et autres
Publié: (2024)

ScienceMeter: Tracking Scientific Knowledge Updates in Language Models
par: Wang, Yike, et autres
Publié: (2025)

Can LLMs Ask Good Questions?
par: Zhang, Yueheng, et autres
Publié: (2025)

Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training
par: Liu, Mingjie, et autres
Publié: (2025)

HREF: Human Response-Guided Evaluation of Instruction Following in Language Models
par: Lyu, Xinxi, et autres
Publié: (2024)

Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning
par: Kim, Joongwon, et autres
Publié: (2024)

WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs
par: Han, Seungju, et autres
Publié: (2024)

The Art of Saying No: Contextual Noncompliance in Language Models
par: Brahman, Faeze, et autres
Publié: (2024)

Strategy Executability in Mathematical Reasoning: Leveraging Human-Model Differences for Effective Guidance
par: Liang, Weida, et autres
Publié: (2026)

EvalTree: Profiling Language Model Weaknesses via Hierarchical Capability Trees
par: Zeng, Zhiyuan, et autres
Publié: (2025)

MentorCollab: Selective Large-to-Small Inference-Time Guidance for Efficient Reasoning
par: Wang, Haojin, et autres
Publié: (2026)

ToRL: Scaling Tool-Integrated RL
par: Li, Xuefeng, et autres
Publié: (2025)

Train for Truth, Keep the Skills: Binary Retrieval-Augmented Reward Mitigates Hallucinations
par: Chen, Tong, et autres
Publié: (2025)

Set the Clock: Temporal Alignment of Pretrained Language Models
par: Zhao, Bowen, et autres
Publié: (2024)

Infini-gram mini: Exact n-gram Search at the Internet Scale with FM-Index
par: Xu, Hao, et autres
Publié: (2025)

RAG-RL: Advancing Retrieval-Augmented Generation via RL and Curriculum Learning
par: Huang, Jerry, et autres
Publié: (2025)

SAIL-RL: Guiding MLLMs in When and How to Think via Dual-Reward RL Tuning
par: Shu, Fangxun, et autres
Publié: (2025)

Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokens
par: Liu, Jiacheng, et autres
Publié: (2024)

Exploiting Tree Structure for Credit Assignment in RL Training of LLMs
par: Tran, Hieu, et autres
Publié: (2025)

WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild
par: Lin, Bill Yuchen, et autres
Publié: (2024)

Organize the Web: Constructing Domains Enhances Pre-Training Data Curation
par: Wettig, Alexander, et autres
Publié: (2025)

Steering off Course: Reliability Challenges in Steering Language Models
par: Da Silva, Patrick Queiroz, et autres
Publié: (2025)

ComPO: Community Preferences for Language Model Personalization
par: Kumar, Sachin, et autres
Publié: (2024)

Small Reward Models via Backward Inference
par: Wang, Yike, et autres
Publié: (2026)

ReviewRL: Towards Automated Scientific Review with RL
par: Zeng, Sihang, et autres
Publié: (2025)

Grokking or Glitching? How Low-Precision Drives Slingshot Loss Spikes
par: Hanqing, Liu, et autres
Publié: (2026)