Cover Image

More Thinking, More Bias: Length-Driven Position Bias in Reasoning Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Author:	Wang, Xiao
Format:	Preprint
Published:	2026
Subjects:	Artificial Intelligence Computation and Language Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

When More is Less: Understanding Chain-of-Thought Length in LLMs
by: Wu, Yuyang, et al.
Published: (2025)

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think
by: Hammoud, Hasan Abed Al Kader, et al.
Published: (2025)

Token Homogenization under Positional Bias
by: Yusupov, Viacheslav, et al.
Published: (2025)

CoDAR: Continuous Diffusion Language Models are More Powerful Than You Think
by: Shen, Junzhe, et al.
Published: (2026)

More Bang for the Buck: Process Reward Modeling with Entropy-Driven Uncertainty
by: Cao, Lang, et al.
Published: (2025)

Revisiting Uncertainty Quantification Evaluation in Language Models: Spurious Interactions with Response Length Bias Results
by: Santilli, Andrea, et al.
Published: (2025)

On the Inductive Bias of Stacking Towards Improving Reasoning
by: Saunshi, Nikunj, et al.
Published: (2024)

Lost in the Middle at Birth: An Exact Theory of Transformer Position Bias
by: Chowdhury, Borun D
Published: (2026)

When Models Know More Than They Say: Probing Analogical Reasoning in LLMs
by: McGovern, Hope, et al.
Published: (2026)

Position-Aware Parameter Efficient Fine-Tuning Approach for Reducing Positional Bias in LLMs
by: Zhang, Zheng, et al.
Published: (2024)

Reasoning Towards Fairness: Mitigating Bias in Language Models through Reasoning-Guided Fine-Tuning
by: Kabra, Sanchit, et al.
Published: (2025)

Mitigating Bias for Question Answering Models by Tracking Bias Influence
by: Ma, Mingyu Derek, et al.
Published: (2023)

Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization
by: Hsieh, Cheng-Yu, et al.
Published: (2024)

DLER: Doing Length pEnalty Right - Incentivizing More Intelligence per Token via Reinforcement Learning
by: Liu, Shih-Yang, et al.
Published: (2025)

More is More: Addition Bias in Large Language Models
by: Santagata, Luca, et al.
Published: (2024)

Parallel LLM Reasoning for Bias-Resilient, Robust Conceptual Abstraction
by: Adeseye, Aisvarya, et al.
Published: (2026)

Anchored Answers: Unravelling Positional Bias in GPT-2's Multiple-Choice Questions
by: Li, Ruizhe, et al.
Published: (2024)

Relative Bias: A Comparative Framework for Quantifying Bias in LLMs
by: Arbabi, Alireza, et al.
Published: (2025)

To Think or Not to Think: Exploring the Unthinking Vulnerability in Large Reasoning Models
by: Zhu, Zihao, et al.
Published: (2025)

AdapThink: Adaptive Thinking Preferences for Reasoning Language Model
by: Wan, Xu, et al.
Published: (2025)

Less Noise, More Voice: Reinforcement Learning for Reasoning via Instruction Purification
by: Guo, Yiju, et al.
Published: (2026)

Towards More Accurate US Presidential Election via Multi-step Reasoning with Large Language Models
by: Yu, Chenxiao, et al.
Published: (2024)

Does Differential Privacy Impact Bias in Pretrained NLP Models?
by: Islam, Md. Khairul, et al.
Published: (2024)

More Expressive Attention with Negative Weights
by: Lv, Ang, et al.
Published: (2024)

Understanding and Mitigating Tokenization Bias in Language Models
by: Phan, Buu, et al.
Published: (2024)

Bias Amplification in Language Model Evolution: An Iterated Learning Perspective
by: Ren, Yi, et al.
Published: (2024)

BiasFreeBench: a Benchmark for Mitigating Bias in Large Language Model Responses
by: Xu, Xin, et al.
Published: (2025)

AdaptThink: Reasoning Models Can Learn When to Think
by: Zhang, Jiajie, et al.
Published: (2025)

Position as Probability: Self-Supervised Transformers that Think Past Their Training for Length Extrapolation
by: Lee, Philip Heejun
Published: (2025)

Packing Analysis: Packing Is More Appropriate for Large Models or Datasets in Supervised Fine-tuning
by: Wang, Shuhe, et al.
Published: (2024)

From Noisy Traces to Stable Gradients: Bias-Variance Optimized Preference Optimization for Aligning Large Reasoning Models
by: Zhu, Mingkang, et al.
Published: (2025)

Less is More for Improving Automatic Evaluation of Factual Consistency
by: Wang, Tong, et al.
Published: (2024)

Causal Estimation of Tokenisation Bias
by: Lesci, Pietro, et al.
Published: (2025)

Less is More: Local Intrinsic Dimensions of Contextual Language Models
by: Ruppik, Benjamin Matthias, et al.
Published: (2025)

More Compute Is What You Need
by: Guo, Zhen
Published: (2024)

LIMR: Less is More for RL Scaling
by: Li, Xuefeng, et al.
Published: (2025)

More Agents Is All You Need
by: Li, Junyou, et al.
Published: (2024)

Thinking About Thinking: SAGE-nano's Inverse Reasoning for Self-Aware Language Models
by: Jha, Basab, et al.
Published: (2025)

Efficient Reasoning with Hidden Thinking
by: Shen, Xuan, et al.
Published: (2025)

One Size Does Not Fit All: A Distribution-Aware Sparsification for More Precise Model Merging
by: Luo, Yingfeng, et al.
Published: (2025)