Work

Mitigating Coordinate Prediction Bias from Positional Encoding Failures

Year: 2026

Type: conference-paper

Abstract: While Multimodal Large Language Models (MLLMs) excel at general vision-language tasks, precise coordinate prediction remains a significant challenge, particularly as highresolution inputs cause visual positional encodings (VPEs) to degrade.We demonstrate that these encoding failures do not result in random noise but instead trigger predictable, directional biases, suggesting that models default to internal spatial priors when grounding signals are weak.To counteract this, we introduce Vision-PE Shuffle Guidance (VPSG), a training-free, inference-time correction method.VPSG isolates position-unconditioned tendencies by shuffling VPEs and utilizes this negative evidence to steer digit decoding through a lightweight finite-state machine.Evaluation on the ScreenSpot-Pro benchmark confirms that VPSG effectively rectifies coordinate drift, yielding consistent improvements in localization accuracy across various model scales without any retraining.Our code is available at https://github.com/taoxj2001/VPSG

Source: Findings of the Association for Computational Linguistics: ACL (Findings of ACL)

DOI: 10.18653/v1/2026.findings-acl.1034

Authors: Xingjian Tao, Yiwei Wang, Yujun Cai, Yihong Luo, Kai Han, Jing Tang