---
title: "Can Video LLMs See Through Illusions? — Ohira 2026"
description: "Studies of how vision-language models respond to optical illusions have mostly looked at image LLMs and static pictures."
authors: ["Souto Ohira", "Tosho Hirasawa", "Mamoru Komachi"]
year: 2026
venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2026.lrec-1.730/"
doi: "10.63317/2s4rwea9k5ji"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/ohira-2026-can-video-llms-see-through/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/ohira-2026-can-video-llms-see-through/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/3/) / Can Video LLMs See Through Illusions? Video-Illusion QA Benchmark Dataset

# Can Video LLMs See Through Illusions? Video-Illusion QA Benchmark Dataset

**Authors:** Souto Ohira, Tosho Hirasawa, [Mamoru Komachi](https://md.research.elamur.ai/free/en/authors/mamoru-komachi/) · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.lrec-1.730/) · [PDF](https://aclanthology.org/2026.lrec-1.730.pdf) · [DOI 10.63317/2s4rwea9k5ji](https://doi.org/10.63317/2s4rwea9k5ji) · [OpenAlex](https://openalex.org/W7160046599)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14)

## Summary
Studies of how vision-language models respond to optical illusions have mostly looked at image LLMs and static pictures. Video LLMs remain largely unexamined, even though they are being considered for robotics and autonomous driving. In those settings, being fooled by a motion-based illusion could be dangerous. The authors build VILQA (Video-Illusion QA Benchmark), a question-answering resource of illusion videos whose effects depend on change over time, such as the motion aftereffect. Videos were gathered from YouTube with English and Japanese searches, from two dedicated illusion channels, and from the Illusion Forum. They were then trimmed in time and space to remove captions and explanations. QA pairs probe three abilities: recognizing illusions as humans do, resisting illusions to report physical reality, and reacting to anomalous events the way people do. The authors evaluate several video LLMs, including one fine-tuned on first-person video QA. Some models show human-like illusion perception, and some resist illusions better than humans. First-person fine-tuning partly moves model responses toward human ones. The benchmark supports both safety testing and cognitive-modelling research.

## Key points
- VILQA contains 275 videos: 172 from English YouTube queries, 59 from Japanese queries and 44 from the Illusion Forum. The authors call it the largest benchmark of temporally driven visual illusions.
- It covers 27 video-specific illusion types, including the motion aftereffect, the lilac chaser, Troxler fading, motion-induced blindness, the enigma illusion and the hollow-mask illusion.
- QA is split into a "Discrepancy" category, where perception differs from physical reality, and an "Anomaly" category of unusual events with no perceptual mismatch. These test illusion recognition, illusion robustness and anomaly sensitivity.
- Building on PredNet results reported by Watanabe and colleagues in 2018 and by Kobayashi and colleagues in 2022, the authors hypothesise that fine-tuning on first-person video makes models more human-like. This alignment was only partial.
- The dataset is released at https://github.com/SDS-NLP/VILQA.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{ohira2026can,
  title = {{Can Video LLMs See Through Illusions? Video-Illusion QA Benchmark Dataset}},
  author = {Ohira, Souto and Hirasawa, Tosho and Komachi, Mamoru},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2s4rwea9k5ji},
  url = {https://doi.org/10.63317/2s4rwea9k5ji}
}
```

- [← previous Can LLMs Translate Italy's Language Varieties?](https://md.research.elamur.ai/free/en/papers/signoroni-2026-can-llms-translate-italy-s/)
- [next → Can You Be More Explicit? A Task and Dataset on Explicitations of Implicit Meaning](https://md.research.elamur.ai/free/en/papers/zeidler-2026-can-you-be-more-explicit/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/3/)

## References

27 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Helen E Clark, John A Perrone, and Robert B Isler. 2013. An illusory size–speed bias and railway crossing collisions. Accident Analysis & Prevention, 55:226–231. [find](https://scholar.google.com/scholar?q=Helen+E+Clark%2C+John+A+Perrone%2C+and+Robert+B+Isler%2E+2013%2E+An+illusory+size%E2%80%93speed+bias+and+railway+crossing+collisions%2E+Accident+Analysis+%26+Prevention%2C+55%3A226%E2%80%93231%2E)
2. Aline F Cretenoud, Lukasz Grzeczkowski, Marina Kunchulia, and Michael H Herzog. 2021. Individual differences in the perception of visual illusions are stable across eyes, time, and measurement methods. Journal of vision, 21(5):26–26. [find](https://scholar.google.com/scholar?q=Aline+F+Cretenoud%2C+Lukasz+Grzeczkowski%2C+Marina+Kunchulia%2C+and+Michael+H+Herzog%2E+2021%2E+Individual+differences+in+the+perception+of+visual+illusions+are+stable+across+eyes%2C+time%2C+and+measurement+methods)
3. R. H. Day. 1984. The nature of, perceptual illusions. Interdisciplinary Science Reviews, 9(1):47–58. [find](https://scholar.google.com/scholar?q=R%2E+H%2E+Day%2E+1984%2E+The+nature+of%2C+perceptual+illusions%2E+Interdisciplinary+Science+Reviews%2C+9%281%29%3A47%E2%80%9358%2E)
4. Bo Dong, Airui Chen, Yuting Zhang, Yangyang Zhang, Ming Zhang, and Tianyang Zhang. 2021. The foggy effect of egocentric distance in a nonverbal paradigm. Scientific Reports, 11(1):14398. [find](https://scholar.google.com/scholar?q=Bo+Dong%2C+Airui+Chen%2C+Yuting+Zhang%2C+Yangyang+Zhang%2C+Ming+Zhang%2C+and+Tianyang+Zhang%2E+2021%2E+The+foggy+effect+of+egocentric+distance+in+a+nonverbal+paradigm%2E+Scientific+Reports%2C+11%281%29%3A14398%2E)
5. Vebjørn Ekroll, Mats Svalebjørg, Angelo Pirrone, Gisela Böhm, Sebastian Jentschke, Rob van Lier, Johan Wagemans, and Alena Høye. 2021. The illusion of absence: how a common feature of magic shows can explain a class of road accidents. Cognitive research: principles and implications, 6:1–16. [find](https://scholar.google.com/scholar?q=Vebj%C3%B8rn+Ekroll%2C+Mats+Svalebj%C3%B8rg%2C+Angelo+Pirrone%2C+Gisela+B%C3%B6hm%2C+Sebastian+Jentschke%2C+Rob+van+Lier%2C+Johan+Wagemans%2C+and+Alena+H%C3%B8ye%2E+2021%2E+The+illusion+of+absence%3A+how+a+common+feature+of+magic+shows+can+)
6. Alexander Gómez-Villa, Adrián Martín, Javier Vazquez-Corral, and Marcelo Bertalmío. 2019. Convolutional neural networks can be deceived by visual illusions. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019, pages 12309–12317. Computer Vi… [find](https://scholar.google.com/scholar?q=Alexander+G%C3%B3mez%2DVilla%2C+Adri%C3%A1n+Mart%C3%ADn%2C+Javier+Vazquez%2DCorral%2C+and+Marcelo+Bertalm%C3%ADo%2E+2019%2E+Convolutional+neural+networks+can+be+deceived+by+visual+illusions%2E+In+IEEE+Conference+on+Computer+Vision+and+P)
7. Alexander Gomez-Villa, Adrian Martín, Javier Vazquez-Corral, Marcelo Bertalmío, and Jesús Malo. 2020. Color illusions also deceive cnns for low-level vision tasks: Analysis and implications. Vision Research, 176:156–174. [find](https://scholar.google.com/scholar?q=Alexander+Gomez%2DVilla%2C+Adrian+Mart%C3%ADn%2C+Javier+Vazquez%2DCorral%2C+Marcelo+Bertalm%C3%ADo%2C+and+Jes%C3%BAs+Malo%2E+2020%2E+Color+illusions+also+deceive+cnns+for+low%2Dlevel+vision+tasks%3A+Analysis+and+implications%2E+Vision+Re)
8. Been Kim, Emily Reif, Martin Wattenberg, Samy Bengio, and Michael C. Mozer. 2021. Neural networks trained on natural scenes exhibit gestalt closure. Computational Brain & Behavior, 4(3):251–263. [find](https://scholar.google.com/scholar?q=Been+Kim%2C+Emily+Reif%2C+Martin+Wattenberg%2C+Samy+Bengio%2C+and+Michael+C%2E+Mozer%2E+2021%2E+Neural+networks+trained+on+natural+scenes+exhibit+gestalt+closure%2E+Computational+Brain+%26+Behavior%2C+4%283%29%3A251%E2%80%93263%2E)
9. Taisuke Kobayashi, Akiyoshi Kitaoka, Manabu Kosaka, Kenta Tanaka, and Eiji Watanabe. 2022. Motion illusion-like patterns extracted from photo and art images using predictive deep neural networks. Scientific Reports, 12(1):3893. [find](https://scholar.google.com/scholar?q=Taisuke+Kobayashi%2C+Akiyoshi+Kitaoka%2C+Manabu+Kosaka%2C+Kenta+Tanaka%2C+and+Eiji+Watanabe%2E+2022%2E+Motion+illusion%2Dlike+patterns+extracted+from+photo+and+art+images+using+predictive+deep+neural+networks%2E+Scie)
10. George Mather, Frans Verstraten, and SM Anstis. 1998. The motion aftereffect: A modern perspective. MIT Press. [find](https://scholar.google.com/scholar?q=George+Mather%2C+Frans+Verstraten%2C+and+SM+Anstis%2E+1998%2E+The+motion+aftereffect%3A+A+modern+perspective%2E+MIT+Press%2E)
11. Donald A Redelmeier and Sheharyar Raza. 2018. Optical illusions and life-threatening traffic crashes: A perspective on aerial perspective. Medical hypotheses, 114:23–27. [find](https://scholar.google.com/scholar?q=Donald+A+Redelmeier+and+Sheharyar+Raza%2E+2018%2E+Optical+illusions+and+life%2Dthreatening+traffic+crashes%3A+A+perspective+on+aerial+perspective%2E+Medical+hypotheses%2C+114%3A23%E2%80%9327%2E)
12. D Samuel Schwarzkopf, Chen Song, and Geraint Rees. 2011. The surface area of human v1 predicts the subjective experience of object size. Nature neuroscience, 14(1):28–30. [find](https://scholar.google.com/scholar?q=D+Samuel+Schwarzkopf%2C+Chen+Song%2C+and+Geraint+Rees%2E+2011%2E+The+surface+area+of+human+v1+predicts+the+subjective+experience+of+object+size%2E+Nature+neuroscience%2C+14%281%29%3A28%E2%80%9330%2E)
13. Taiga Shinozaki, Tomoki Doi, Amane Watahiki, Satoshi Nishida, and Hitomi Yanaka. 2025. Do large vision-language models distinguish between the actual and apparent features of illusions? In Proceedings of the 47th Annual Meeting of the Cognitive Science Society, CogSci 2025. Cognitive Science Society… [find](https://scholar.google.com/scholar?q=Taiga+Shinozaki%2C+Tomoki+Doi%2C+Amane+Watahiki%2C+Satoshi+Nishida%2C+and+Hitomi+Yanaka%2E+2025%2E+Do+large+vision%2Dlanguage+models+distinguish+between+the+actual+and+apparent+features+of+illusions%3F+In+Proceedings)
14. Eric D. Sun and Ron Dekel. 2021. ImageNet-trained deep neural networks exhibit illusion-like response to the Scintillating grid. Journal of Vision, 21(11):15. [find](https://scholar.google.com/scholar?q=Eric+D%2E+Sun+and+Ron+Dekel%2E+2021%2E+ImageNet%2Dtrained+deep+neural+networks+exhibit+illusion%2Dlike+response+to+the+Scintillating+grid%2E+Journal+of+Vision%2C+21%2811%29%3A15%2E)
15. Dejan Todorović. 2020. What are visual illusions? Perception, 49(11):1128–1199. [find](https://scholar.google.com/scholar?q=Dejan+Todorovi%C4%87%2E+2020%2E+What+are+visual+illusions%3F+Perception%2C+49%2811%29%3A1128%E2%80%931199%2E)
16. Eiji Watanabe, Akiyoshi Kitaoka, Kiwako Sakamoto, Masaki Yasugi, and Kenta Tanaka. 2018. Illusory motion reproduced by deep neural networks trained for prediction. Frontiers in psychology, 9:340023. [find](https://scholar.google.com/scholar?q=Eiji+Watanabe%2C+Akiyoshi+Kitaoka%2C+Kiwako+Sakamoto%2C+Masaki+Yasugi%2C+and+Kenta+Tanaka%2E+2018%2E+Illusory+motion+reproduced+by+deep+neural+networks+trained+for+prediction%2E+Frontiers+in+psychology%2C+9%3A340023%2E)
17. Yichi Zhang, Jiayi Pan, Yuchen Zhou, Rui Pan, and Joyce Chai. 2023. Grounding visual illusions in language: Do vision-language models perceive illusions like humans? In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023,… [find](https://scholar.google.com/scholar?q=Yichi+Zhang%2C+Jiayi+Pan%2C+Yuchen+Zhou%2C+Rui+Pan%2C+and+Joyce+Chai%2E+2023%2E+Grounding+visual+illusions+in+language%3A+Do+vision%2Dlanguage+models+perceive+illusions+like+humans%3F+In+Proceedings+of+the+2023+Confere)
18. Language Resource References Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Chen… [find](https://scholar.google.com/scholar?q=Language+Resource+References+Shuai+Bai%2C+Keqin+Chen%2C+Xuejing+Liu%2C+Jialin+Wang%2C+Wenbin+Ge%2C+Sibo+Song%2C+Kai+Dang%2C+Peng+Wang%2C+Shijie+Wang%2C+Jun+Tang%2C+Humen+Zhong%2C+Yuanzhi+Zhu%2C+Mingkun+Yang%2C+Zhaohai+Li%2C+Jian) [our card](https://md.research.elamur.ai/free/en/papers/bai-2025-qwen2-5-vl-technical/)
19. Kaituo Feng, Kaixiong Gong, Bohao Li, Zonghao Guo, Yibing Wang, Tianshuo Peng, Junfei Wu, Xiaoying Zhang, Benyou Wang, and Xiangyu Yue. 2025. Video-R1: Reinforcing video reasoning in MLLMs. [find](https://scholar.google.com/scholar?q=Kaituo+Feng%2C+Kaixiong+Gong%2C+Bohao+Li%2C+Zonghao+Guo%2C+Yibing+Wang%2C+Tianshuo+Peng%2C+Junfei+Wu%2C+Xiaoying+Zhang%2C+Benyou+Wang%2C+and+Xiangyu+Yue%2E+2025%2E+Video%2DR1%3A+Reinforcing+video+reasoning+in+MLLMs%2E)
20. Google. 2025. Gemini 2.5 pro. [https://cloud.google.com/vertex-ai/](https://cloud.google.com/vertex-ai/) generative-ai/docs/models/gemini/ 2-5-pro. Accessed: 2025-6-28. [↗](https://cloud.google.com/vertex-ai/)
21. Tianrui Guan, Fuxiao Liu, Xiyang Wu, Ruiqi Xian, Zongxia Li, Xiaoyu Liu, Xijun Wang, Lichang Chen, Furong Huang, Yaser Yacoob, Dinesh Manocha, and Tianyi Zhou. 2024. Hallusionbench: An advanced diagnostic suite for entangled language hallucination and visual illusion in large vision-language models.… [find](https://scholar.google.com/scholar?q=Tianrui+Guan%2C+Fuxiao+Liu%2C+Xiyang+Wu%2C+Ruiqi+Xian%2C+Zongxia+Li%2C+Xiaoyu+Liu%2C+Xijun+Wang%2C+Lichang+Chen%2C+Furong+Huang%2C+Yaser+Yacoob%2C+Dinesh+Manocha%2C+and+Tianyi+Zhou%2E+2024%2E+Hallusionbench%3A+An+advanced+diagno) [our card](https://md.research.elamur.ai/free/en/papers/guan-2023-hallusionbench-advanced-diagnostic-suite-entangled/)
22. Muhammad Maaz, Hanoona Abdul Rasheed, Salman Khan, and Fahad Khan. 2024. Video-ChatGPT: Towards detailed video understanding via large vision and language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok, Th… [find](https://scholar.google.com/scholar?q=Muhammad+Maaz%2C+Hanoona+Abdul+Rasheed%2C+Salman+Khan%2C+and+Fahad+Khan%2E+2024%2E+Video%2DChatGPT%3A+Towards+detailed+video+understanding+via+large+vision+and+language+models%2E+In+Proceedings+of+the+62nd+Annual+Mee) [our card](https://md.research.elamur.ai/free/en/papers/maaz-2023-video-chatgpt-towards-detailed-video/)
23. Mohammadmostafa Rostamkhani, Baktash Ansari, Hoorieh Sabzevari, Farzan Rahmani, and Sauleh Eetemadi. 2025. Illusory VQA: benchmarking and enhancing multimodal models on visual illusions. In IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPR Workshops 2025, Nashville, TN,… [find](https://scholar.google.com/scholar?q=Mohammadmostafa+Rostamkhani%2C+Baktash+Ansari%2C+Hoorieh+Sabzevari%2C+Farzan+Rahmani%2C+and+Sauleh+Eetemadi%2E+2025%2E+Illusory+VQA%3A+benchmarking+and+enhancing+multimodal+models+on+visual+illusions%2E+In+IEEE%2FCVF+C)
24. Haz Sameen Shahgir, Khondker Salman Sayeed, Abhik Bhattacharjee, Wasi Uddin Ahmad, Yue Dong, and Rifat Shahriyar. 2024. IllusionVQA: A challenging optical illusion dataset for vision language models. [find](https://scholar.google.com/scholar?q=Haz+Sameen+Shahgir%2C+Khondker+Salman+Sayeed%2C+Abhik+Bhattacharjee%2C+Wasi+Uddin+Ahmad%2C+Yue+Dong%2C+and+Rifat+Shahriyar%2E+2024%2E+IllusionVQA%3A+A+challenging+optical+illusion+dataset+for+vision+language+models%2E)
25. CONFERENCE ON LANGUAGE MODELING 2024, COLM 2024. [find](https://scholar.google.com/scholar?q=CONFERENCE+ON+LANGUAGE+MODELING+2024%2C+COLM+2024%2E)
26. Yi Wang, Kunchang Li, Xinhao Li, Jiashuo Yu, Yinan He, Guo Chen, Baoqi Pei, Rongkun Zheng, Zun Wang, Yansong Shi, Tianxiang Jiang, Songze Li, Jilan Xu, Hongjie Zhang, Yifei Huang, Yu Qiao, Yali Wang, and Limin Wang. 2024. InternVideo2: Scaling foundation models for multimodal video understanding. In… [find](https://scholar.google.com/scholar?q=Yi+Wang%2C+Kunchang+Li%2C+Xinhao+Li%2C+Jiashuo+Yu%2C+Yinan+He%2C+Guo+Chen%2C+Baoqi+Pei%2C+Rongkun+Zheng%2C+Zun+Wang%2C+Yansong+Shi%2C+Tianxiang+Jiang%2C+Songze+Li%2C+Jilan+Xu%2C+Hongjie+Zhang%2C+Yifei+Huang%2C+Yu+Qiao%2C+Yali+Wang%2C+)
27. Yuanhan Zhang, Jinming Wu, Wei Li, Bo Li, Zejun Ma, Ziwei Liu, and Chunyuan Li. 2025. LLaVA-video: Video instruction tuning with synthetic data. Transactions on Machine Learning Research. [find](https://scholar.google.com/scholar?q=Yuanhan+Zhang%2C+Jinming+Wu%2C+Wei+Li%2C+Bo+Li%2C+Zejun+Ma%2C+Ziwei+Liu%2C+and+Chunyuan+Li%2E+2025%2E+LLaVA%2Dvideo%3A+Video+instruction+tuning+with+synthetic+data%2E+Transactions+on+Machine+Learning+Research%2E)