Quantification and object perception in Multimodal Large Language Models deviate from human linguistic cognition

2025-11-11

AI Summary

This study investigates systematic cognitive biases in multimodal large language models (MLLMs) regarding quantifier representation, focusing on three cross-linguistically robust features of human quantitative cognition: quantifier ordering scales, usage scope/typicality, and Approximate Number System (ANS)-related biases. Method: Leveraging psycholinguistic paradigms and multilingual comparative experiments, we design a quantification comprehension evaluation task—introducing, for the first time, a three-dimensional human cognitive framework into MLLM assessment. Contribution/Results: We find that MLLMs exhibit systematic deviations in both semantic and pragmatic quantifier representations. These biases are significantly modulated by model architecture (e.g., presence of visual grounding capabilities) and linguistic typology (e.g., classifier richness, numeral system properties). Our findings reveal fundamental limitations in current MLLMs’ numerical reasoning and provide both theoretical grounding and empirical benchmarks for developing cognitively aligned multimodal quantification models.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Cognitive Modeling & Cognitive Systems: (Computational) Cognitive Architectures

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Quantification has been proven to be a particularly difficult linguistic phenomenon for (Multimodal) Large Language Models (MLLMs). However, given that quantification interfaces with the logic, pragmatic, and numerical domains, the exact reasons for the poor performance are still unclear. This papers looks at three key features of human quantification shared cross-linguistically that have remained so far unexplored in the (M)LLM literature: the ordering of quantifiers into scales, the ranges of use and prototypicality, and the biases inherent in the human approximate number system. The aim is to determine how these features are encoded in the models'architecture, how they may differ from humans, and whether the results are affected by the type of model and language under investigation. We find that there are clear differences between humans and MLLMs with respect to these features across various tasks that tap into the representation of quantification in vivo vs. in silico. This work, thus, paves the way for addressing the nature of MLLMs as semantic and pragmatic agents, while the cross-linguistic lens can elucidate whether their abilities are robust and stable across different languages.

Problem

• Investigates MLLMs' deviation from human quantification cognition across languages
• Analyzes scaling, prototypicality and numerical biases in multimodal language models
• Examines cross-linguistic robustness of semantic representations in AI systems

Innovation

• Analyzes quantifier scales and prototypicality in models
• Compares human vs model quantification across languages
• Examines architectural encoding of semantic biases