Cover Image

Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ren, Qihan, Wang, Peng, Cai, Ruikun, Shao, Shuai, Guo, Dadi, Xie, Yuejin, Li, Yafu, Zhang, Quanshi, Hu, Xia, Shao, Jing, Liu, Dongrui
Format:	Preprint
Published:	2026
Subjects:	Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

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

Rethinking Entropy Regularization in Large Reasoning Models
by: Jiang, Yuxian, et al.
Published: (2025)

Attributing Emergence in Million-Agent Systems
by: Tang, Ling, et al.
Published: (2026)

The Why Behind the Action: Unveiling Internal Drivers via Agentic Attribution
by: Qian, Chen, et al.
Published: (2026)

Your Agent May Misevolve: Emergent Risks in Self-evolving LLM Agents
by: Shao, Shuai, et al.
Published: (2025)

ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis
by: Li, Yu, et al.
Published: (2026)

Towards Self-Evolving Benchmarks: Synthesizing Agent Trajectories via Test-Time Exploration under Validate-by-Reproduce Paradigm
by: Guo, Dadi, et al.
Published: (2025)

Towards the Dynamics of a DNN Learning Symbolic Interactions
by: Ren, Qihan, et al.
Published: (2024)

Code2Math: Can Your Code Agent Effectively Evolve Math Problems Through Exploration?
by: Guo, Dadi, et al.
Published: (2026)

RiOSWorld: Benchmarking the Risk of Multimodal Computer-Use Agents
by: Yang, Jingyi, et al.
Published: (2025)

Are Your Agents Upward Deceivers?
by: Guo, Dadi, et al.
Published: (2025)

ReasonAny: Incorporating Reasoning Capability to Any Model via Simple and Effective Model Merging
by: Yang, Junyao, et al.
Published: (2026)

Where We Have Arrived in Proving the Emergence of Sparse Symbolic Concepts in AI Models
by: Ren, Qihan, et al.
Published: (2023)

Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions
by: Cheng, Lei, et al.
Published: (2025)

Evaluating the Correctness of Inference Patterns Used by LLMs for Judgment
by: Chen, Lu, et al.
Published: (2024)

Benchmarks for Trajectory Safety Evaluation and Diagnosis in OpenClaw and Codex: ATBench-Claw and ATBench-Codex
by: Yang, Zhonghao, et al.
Published: (2026)

ExGRPO: Learning to Reason from Experience
by: Zhan, Runzhe, et al.
Published: (2025)

Reconciling Contradictory Views on the Effectiveness of SFT in LLMs: An Interaction Perspective
by: Zhang, Junpeng, et al.
Published: (2026)

What Do EEG Foundation Models Capture from Human Brain Signals?
by: Tang, Ling, et al.
Published: (2026)

COLLEAGUE.SKILL: Automated AI Skill Generation via Expert Knowledge Distillation
by: Zhou, Tianyi, et al.
Published: (2026)

Independent characterization of the elastic and the mixing parts of hydrogel osmotic pressure
by: Shao, Zefan, et al.
Published: (2023)

PACEbench: A Framework for Evaluating Practical AI Cyber-Exploitation Capabilities
by: Liu, Zicheng, et al.
Published: (2025)

PRISM: Preference-Aware Influence Function Based Data Selection Method for Efficient Fine-Tuning
by: Lin, Qihao, et al.
Published: (2026)

Loop as a Bridge: Can Looped Transformers Truly Link Representation Space and Natural Language Outputs?
by: Chen, Guanxu, et al.
Published: (2026)

Learning to Adapt SFT Data for Better Reasoning Generalization
by: Sun, Lisong, et al.
Published: (2026)

Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report v1.5
by: Liu, Dongrui, et al.
Published: (2026)

Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Deliberation
by: Zhang, Haoran, et al.
Published: (2025)

Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning
by: Qian, Chen, et al.
Published: (2025)

Identifying Semantic Induction Heads to Understand In-Context Learning
by: Ren, Jie, et al.
Published: (2024)

Rethinking Prospect Theory for LLMs: Revealing the Instability of Decision-Making under Epistemic Uncertainty
by: Wang, Rui, et al.
Published: (2025)

AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security
by: Liu, Dongrui, et al.
Published: (2026)

Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models
by: Yang, Junyao, et al.
Published: (2026)

LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
by: Hu, Xuhao, et al.
Published: (2025)

The Interaction Bottleneck of Deep Neural Networks: Discovery, Proof, and Modulation
by: Deng, Huiqi, et al.
Published: (2025)

Memorize Theorems, Not Instances: Probing SFT Generalization through Mathematical Reasoning
by: Peng, Ruiying, et al.
Published: (2026)

VLSBench: Unveiling Visual Leakage in Multimodal Safety
by: Hu, Xuhao, et al.
Published: (2024)

PrivacyPeek: Auditing What LLM-Based Agents Acquire, Not Just What They Say
by: Zhang, Mingxuan, et al.
Published: (2026)

Tree-frog-inspired osmocapillary adhesive bonding to diverse substrates
by: Shao, Zefan, et al.
Published: (2025)

Interpreting Emergent Extreme Events in Multi-Agent Systems
by: Tang, Ling, et al.
Published: (2026)

Bridging SFT and RL: Dynamic Policy Optimization for Robust Reasoning
by: Zhu, Taojie, et al.
Published: (2026)