Buchumschlag

The First Impression Problem: Internal Bias Triggers Overthinking in Reasoning Models

Fuente: arXiv

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Dang, Renfei, Li, Zhening, Huang, Shujian, Chen, Jiajun
Format:	Preprint
Veröffentlicht:	2025
Schlagworte:	Artificial Intelligence
Online-Zugang:	Acceder al recurso
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Ähnliche Einträge

Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoning
von: Hong, Jialiang, et al.
Veröffentlicht: (2025)

BadThink: Triggered Overthinking Attacks on Chain-of-Thought Reasoning in Large Language Models
von: Liu, Shuaitong, et al.
Veröffentlicht: (2025)

Mitigating Overthinking through Reasoning Shaping
von: Song, Feifan, et al.
Veröffentlicht: (2025)

Measuring Meaning Composition in the Human Brain with Composition Scores from Large Language Models
von: Gao, Changjiang, et al.
Veröffentlicht: (2024)

Mitigating Overthinking in Large Reasoning Models via Manifold Steering
von: Huang, Yao, et al.
Veröffentlicht: (2025)

Innate Reasoning is Not Enough: In-Context Learning Enhances Reasoning Large Language Models with Less Overthinking
von: Ge, Yuyao, et al.
Veröffentlicht: (2025)

Mitigating Overthinking in Large Reasoning Language Models via Reasoning Path Deviation Monitoring
von: Guan, Weixin, et al.
Veröffentlicht: (2026)

The Shape of Overthinking: Backtracking Bursts in Long Reasoning Traces
von: Rezazadeh, Navid, et al.
Veröffentlicht: (2026)

Think How to Think: Mitigating Overthinking with Autonomous Difficulty Cognition in Large Reasoning Models
von: Liu, Yongjiang, et al.
Veröffentlicht: (2025)

The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks
von: Cuadron, Alejandro, et al.
Veröffentlicht: (2025)

Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Models
von: Yue, Linan, et al.
Veröffentlicht: (2025)

Precedent-Informed Reasoning: Mitigating Overthinking in Large Reasoning Models via Test-Time Precedent Learning
von: Wang, Qianyue, et al.
Veröffentlicht: (2026)

Your Models Have Thought Enough: Training Large Reasoning Models to Stop Overthinking
von: Han, Jinyi, et al.
Veröffentlicht: (2025)

Reasoning or Overthinking: Evaluating Large Language Models on Financial Sentiment Analysis
von: Vamvourellis, Dimitris, et al.
Veröffentlicht: (2025)

Missing Premise exacerbates Overthinking: Are Reasoning Models losing Critical Thinking Skill?
von: Fan, Chenrui, et al.
Veröffentlicht: (2025)

Mitigating Overthinking in Large Reasoning Models via Difficulty-aware Reinforcement Learning
von: Wan, Qian, et al.
Veröffentlicht: (2026)

Lost in the Source Language: How Large Language Models Evaluate the Quality of Machine Translation
von: Huang, Xu, et al.
Veröffentlicht: (2024)

The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis
von: Wei, Zihao, et al.
Veröffentlicht: (2025)

Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoning
von: Hassid, Michael, et al.
Veröffentlicht: (2025)

Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMs
von: Su, Jinyan, et al.
Veröffentlicht: (2025)

Hop, Skip, and Overthink: Diagnosing Why Reasoning Models Fumble during Multi-Hop Analysis
von: Yadav, Anushka, et al.
Veröffentlicht: (2025)

Don't "Overthink" Passage Reranking: Is Reasoning Truly Necessary?
von: Jedidi, Nour, et al.
Veröffentlicht: (2025)

From Token to Action: State Machine Reasoning to Mitigate Overthinking in Information Retrieval
von: Lee, Dohyeon, et al.
Veröffentlicht: (2025)

TACT: Mitigating Overthinking and Overacting in Coding Agents via Activation Steering
von: Sui, Yuan, et al.
Veröffentlicht: (2026)

POT: Inducing Overthinking in LLMs via Black-Box Iterative Optimization
von: Li, Xinyu, et al.
Veröffentlicht: (2025)

Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models
von: Wang, Shuqiang, et al.
Veröffentlicht: (2026)

Reasoning While Asking: Transforming Reasoning Large Language Models from Passive Solvers to Proactive Inquirers
von: Chen, Xin, et al.
Veröffentlicht: (2026)

When More Thinking Hurts: Overthinking in LLM Test-Time Compute Scaling
von: Zhou, Shu, et al.
Veröffentlicht: (2026)

TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models
von: Zhao, Yi, et al.
Veröffentlicht: (2026)

Can Large Models Teach Student Models to Solve Mathematical Problems Like Human Beings? A Reasoning Distillation Method via Multi-LoRA Interaction
von: Li, Xinhe, et al.
Veröffentlicht: (2025)

Model-First Reasoning LLM Agents: Reducing Hallucinations through Explicit Problem Modeling
von: Rana, Annu, et al.
Veröffentlicht: (2025)

Internalizing Safety Understanding in Large Reasoning Models via Verification
von: Zhang, Yi, et al.
Veröffentlicht: (2026)

Generalizing From Short to Long: Effective Data Synthesis for Long-Context Instruction Tuning
von: Zhu, Wenhao, et al.
Veröffentlicht: (2025)

Calibrating Reasoning in Language Models with Internal Consistency
von: Xie, Zhihui, et al.
Veröffentlicht: (2024)

Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality?
von: Kang, Caixin, et al.
Veröffentlicht: (2026)

Overthinking Reduction with Decoupled Rewards and Curriculum Data Scheduling
von: Jiang, Shuyang, et al.
Veröffentlicht: (2025)

Enhancing Mathematical Reasoning in LLMs with Background Operators
von: Chen, Jiajun, et al.
Veröffentlicht: (2024)

Overthinking the Truth: Understanding how Language Models Process False Demonstrations
von: Halawi, Danny, et al.
Veröffentlicht: (2023)

Learning to Stop Overthinking at Test Time
von: Bao, Hieu Tran, et al.
Veröffentlicht: (2025)

Making Mathematical Reasoning Adaptive
von: Lai, Zhejian, et al.
Veröffentlicht: (2025)