Do Multimodal LLMs Understand Order? Measuring the Fragility of Multimodal Reasoning under Input Order Perturbations

Sheng-Lun Wei, Yu-Ling Liao, Hen-Hsen Huang, Hsin-Hsi Chen


Abstract
Multimodal reasoning has progressed rapidly with large vision-language models (LVLMs), yet their robustness under input variations remains underexplored. This study investigates positional bias in LVLMs for multimodal multiple-choice questions. Our analysis shows that model predictions are sensitive to both choice and modality ordering. We conduct a large-scale evaluation on MMMU, CVQA, and MMBench using fourteen representative models. Further analysis examines how question properties, including difficulty, domain, and image type, affect robustness. We also assess whether text-based mitigation strategies transfer to the VQA setting and perform ablation studies on self-consistency and reasoning complexity. Overall, our findings provide the first comprehensive understanding of positional bias from a vision-language perspective, highlighting key challenges in achieving stable multimodal reasoning.
Anthology ID:
2026.lrec-1.716
Volume:
Proceedings of the Fifteenth Language Resources and Evaluation Conference
Month:
May
Year:
2026
Address:
Palma de Mallorca, Spain
Editors:
Stelios Piperidis, Núria Bel, Henk van den Heuvel, Nancy Ide, Simon Krek, Antonio Toral
Venue:
LREC
SIG:
Publisher:
ELRA Language Resource Association
Note:
Pages:
9118–9128
Language:
External URL:
https://lrec.elra.info/lrec2026-main-716
DOI:
10.63317/4jtpgzks8pbr
Bibkey:
Cite (ACL):
Sheng-Lun Wei, Yu-Ling Liao, Hen-Hsen Huang, and Hsin-Hsi Chen. 2026. Do Multimodal LLMs Understand Order? Measuring the Fragility of Multimodal Reasoning under Input Order Perturbations. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 9118–9128, Palma de Mallorca, Spain. ELRA Language Resource Association.
Cite (Informal):
Do Multimodal LLMs Understand Order? Measuring the Fragility of Multimodal Reasoning under Input Order Perturbations (Wei et al., LREC 2026)
Copy Citation: