Cognitive Mismatch in Multimodal Large Language Models for Discrete Symbol Understanding
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908948871774208 |
|---|---|
| author | Li, Yinghui Kuang, Jiayi Xing, Peng Liu, Daixian Zhang, Yongheng Dong, Junnan Guo, Shu-Yu Li, Yangning Zhou, Qingyu Jiang, Wenhao Zheng, Hai-Tao Shen, Ying Lin, Liang Yu, Philip S. |
| author_facet | Li, Yinghui Kuang, Jiayi Xing, Peng Liu, Daixian Zhang, Yongheng Dong, Junnan Guo, Shu-Yu Li, Yangning Zhou, Qingyu Jiang, Wenhao Zheng, Hai-Tao Shen, Ying Lin, Liang Yu, Philip S. |
| contents | Multimodal large language models (MLLMs) perform strongly on natural images, yet their ability to understand discrete visual symbols remains unclear. We present a multi-domain benchmark spanning language, culture, mathematics, physics and chemistry, organized into three cognitive levels: perception and recognition, combination and reasoning, and association and critical thinking. Across leading MLLMs, we observe a consistent cognitive mismatch. Models frequently underperform on elementary symbol recognition while appearing relatively competent on more complex reasoning tasks. This recognition-reasoning inversion indicates that current systems often compensate with linguistic priors, template retrieval or procedural reasoning instead of robust visual grounding. The pattern is especially clear for sparse, low-redundancy symbols such as handwritten characters, formula graphs, circuit diagrams and chemical structures. These results show that symbolic understanding remains a major bottleneck for multimodal intelligence and motivate training and evaluation schemes that prioritize grounded perception in discrete semantic spaces. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_18472 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Cognitive Mismatch in Multimodal Large Language Models for Discrete Symbol Understanding Li, Yinghui Kuang, Jiayi Xing, Peng Liu, Daixian Zhang, Yongheng Dong, Junnan Guo, Shu-Yu Li, Yangning Zhou, Qingyu Jiang, Wenhao Zheng, Hai-Tao Shen, Ying Lin, Liang Yu, Philip S. Artificial Intelligence Computer Vision and Pattern Recognition Multimodal large language models (MLLMs) perform strongly on natural images, yet their ability to understand discrete visual symbols remains unclear. We present a multi-domain benchmark spanning language, culture, mathematics, physics and chemistry, organized into three cognitive levels: perception and recognition, combination and reasoning, and association and critical thinking. Across leading MLLMs, we observe a consistent cognitive mismatch. Models frequently underperform on elementary symbol recognition while appearing relatively competent on more complex reasoning tasks. This recognition-reasoning inversion indicates that current systems often compensate with linguistic priors, template retrieval or procedural reasoning instead of robust visual grounding. The pattern is especially clear for sparse, low-redundancy symbols such as handwritten characters, formula graphs, circuit diagrams and chemical structures. These results show that symbolic understanding remains a major bottleneck for multimodal intelligence and motivate training and evaluation schemes that prioritize grounded perception in discrete semantic spaces. |
| title | Cognitive Mismatch in Multimodal Large Language Models for Discrete Symbol Understanding |
| topic | Artificial Intelligence Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.18472 |