VLMs May Not Globally Enhance Human Alignment over LLMs During Natural Reading
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914609119625216 |
|---|---|
| author | Wu, Jinzhou Ma, Zhengwu Li, Jixing Tang, Baoping Lu, Zitong |
| author_facet | Wu, Jinzhou Ma, Zhengwu Li, Jixing Tang, Baoping Lu, Zitong |
| contents | Large language models (LLMs) have become increasingly useful computational models of human language processing, but it remains unclear whether vision-language learning makes text representations more human-like during natural reading. Here, we address this question by comparing tightly matched LLM and vision-language model (VLM) pairs under a strictly text-only setting, allowing us to isolate the effect of multimodal training history from online visual input or cross-modal fusion. We evaluate model alignment with a human natural-reading dataset that includes whole-cortex fMRI responses and synchronized eye-tracking saccades. Our findings demonstrate that multimodal pretraining may not confer a uniform, global advantage in human alignment during natural reading, indicating that language-internal representations remain the key factor for modeling human text processing. However, the VLM advantage could emerge more selectively when sentences contain stronger visual semantic content, with converging evidence from both fMRI and eye-movement alignments. Together, our findings provide a controlled in silico framework for testing how visual learning history shapes model-human alignment of language processing, suggesting that multimodal pretraining contributes selectively rather than globally to human-like language representations during natural reading. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_28818 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | VLMs May Not Globally Enhance Human Alignment over LLMs During Natural Reading Wu, Jinzhou Ma, Zhengwu Li, Jixing Tang, Baoping Lu, Zitong Computation and Language Neurons and Cognition Large language models (LLMs) have become increasingly useful computational models of human language processing, but it remains unclear whether vision-language learning makes text representations more human-like during natural reading. Here, we address this question by comparing tightly matched LLM and vision-language model (VLM) pairs under a strictly text-only setting, allowing us to isolate the effect of multimodal training history from online visual input or cross-modal fusion. We evaluate model alignment with a human natural-reading dataset that includes whole-cortex fMRI responses and synchronized eye-tracking saccades. Our findings demonstrate that multimodal pretraining may not confer a uniform, global advantage in human alignment during natural reading, indicating that language-internal representations remain the key factor for modeling human text processing. However, the VLM advantage could emerge more selectively when sentences contain stronger visual semantic content, with converging evidence from both fMRI and eye-movement alignments. Together, our findings provide a controlled in silico framework for testing how visual learning history shapes model-human alignment of language processing, suggesting that multimodal pretraining contributes selectively rather than globally to human-like language representations during natural reading. |
| title | VLMs May Not Globally Enhance Human Alignment over LLMs During Natural Reading |
| topic | Computation and Language Neurons and Cognition |
| url | https://arxiv.org/abs/2605.28818 |