Publications

@article{sani2026imagenworld,
  title={ImagenWorld: Stress-Testing Image Generation Models with Explainable Human Evaluation on Open-ended Real-World Tasks},
  author={Sani, Samin Mahdizadeh and Ku, Max and Jamali, Nima and Sani, Matina Mahdizadeh and Khoshtab, Paria and Sun, Wei-Chieh and Fazel, Parnian and Tam, Zhi Rui and Chong, Thomas and Chan, Edisy Kin Wai and others},
  journal={arXiv preprint arXiv:2603.27862},
  year={2026}
}

@article{huang2025five,
  title={Five Years of SciCap: What We Learned and Future Directions for Scientific Figure Captioning},
  author={Huang, Ting-Hao'Kenneth' and Rossi, Ryan A and Kim, Sungchul and Yu, Tong and Hsu, Ting-Yao E and Yin, Ho and Giles, C Lee and others},
  journal={arXiv preprint arXiv:2512.21789},
  year={2025}
}

@article{10.1162/TACL.a.653,
  author = {Hsu, Ting-Yao ‘Edward’ and Hsu, Yi-Li and Rohatgi, Shaurya and Huang, Chieh-Yang and Ng, Ho Yin Sam and Rossi, Ryan and Kim, Sungchul and Yu, Tong and Ku, Lun-Wei and Giles, Clyde Lee and Huang, Ting-Hao ‘Kenneth’},
  title = {Do Large Multimodal Models Solve Caption Generation for Scientific Figures? Lessons Learned from SciCap Challenge 2023},
  journal = {Transactions of the Association for Computational Linguistics},
  volume = {14},
  pages = {233-252},
  year = {2026},
  month = {03},
  abstract = {Since the SciCap dataset’s launch in 2021, the research community has made significant progress in generating captions for scientific figures in scholarly articles. In 2023, the first SciCap Challenge took place, inviting global teams to use an expanded SciCap dataset to develop models for captioning diverse figure types across various academic fields. At the same time, text generation models advanced quickly, with many powerful pre-trained large multimodal models (LMMs) emerging that showed impressive capabilities in various vision-and-language tasks. This paper presents an overview of the first SciCap Challenge and details the performance of various models on its data, capturing a snapshot of the field’s state. We found that professional editors overwhelmingly preferred figure captions generated by GPT-4V over those from all other models and even the original captions written by authors. Following this key finding, we conducted detailed analyses to answer this question: Have advanced LMMs solved the task of generating captions for scientific figures?},
  issn = {2307-387X},
  doi = {10.1162/TACL.a.653},
  url = {https://doi.org/10.1162/TACL.a.653},
  eprint = {https://direct.mit.edu/tacl/article-pdf/doi/10.1162/TACL.a.653/2587241/tacl.a.653.pdf},
}

@inproceedings{ng-etal-2025-lamp,
title = "{L}a{MP}-Cap: Personalized Figure Caption Generation With Multimodal Figure Profiles",
author = "Ng, Ho Yin Sam  and
  Hsu, Edward  and
  Anantha Ramakrishnan, Aashish  and
  Kveton, Branislav  and
  Lipka, Nedim  and
  Dernoncourt, Franck  and
  Lee, Dongwon  and
  Yu, Tong  and
  Kim, Sungchul  and
  Rossi, Ryan A.  and
  Huang, Ting-Hao Kenneth",
editor = "Christodoulopoulos, Christos  and
  Chakraborty, Tanmoy  and
  Rose, Carolyn  and
  Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.521/",
doi = "10.18653/v1/2025.findings-emnlp.521",
pages = "9818--9832",
ISBN = "979-8-89176-335-7",
abstract = "Figure captions are crucial for helping readers understand and remember a figure{'}s key message. Many models have been developed to generate these captions, helping authors compose better quality captions more easily. Yet, authors almost always need to revise generic AI-generated captions to match their writing style and the domain{'}s style, highlighting the need for personalization. Despite language models' personalization (LaMP) advances, these technologies often focus on text-only settings and rarely address scenarios where both inputs and profiles are multimodal. This paper introduces LaMP-Cap, a dataset for personalized figure caption generation with multimodal figure profiles. For each target figure, LaMP-Cap provides not only the needed inputs, such as figure images, but also up to three other figures from the same document{---}each with its image, caption, and figure-mentioning paragraphs{---}as a profile to characterize the context. Experiments with four LLMs show that using profile information consistently helps generate captions closer to the original author-written ones. Ablation studies reveal that images in the profile are more helpful than figure-mentioning paragraphs, highlighting the advantage of using multimodal profiles over text-only ones."
}

@inproceedings{ng-etal-2025-understanding,
title = "Understanding Writing Assistants for Scientific Figure Captions: A Thematic Analysis",
author = "Ng, Ho Yin Sam  and
  Hsu, Ting-Yao  and
  Min, Jiyoo  and
  Kim, Sungchul  and
  Rossi, Ryan A.  and
  Yu, Tong  and
  Jung, Hyunggu  and
  Huang, Ting-Hao Kenneth",
editor = "Padmakumar, Vishakh  and
  Gero, Katy  and
  Wambsganss, Thiemo  and
  Sterman, Sarah  and
  Huang, Ting-Hao  and
  Zhou, David  and
  Chung, John",
booktitle = "Proceedings of the Fourth Workshop on Intelligent and Interactive Writing Assistants (In2Writing 2025)",
month = may,
year = "2025",
address = "Albuquerque, New Mexico, US",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.in2writing-1.1/",
doi = "10.18653/v1/2025.in2writing-1.1",
pages = "1--10",
ISBN = "979-8-89176-239-8",
abstract = "Scientific figure captions are essential for communicating complex data but are often overlooked, leading to unclear or redundant descriptions. While many studies focus on generating captions as an `output', little attention has been given to the writer{'}s process of crafting captions for scientific figures. This study examines how researchers use AI-generated captions to support caption writing. Through thematic analysis of interviews and video recordings with 18 participants from diverse disciplines, we identified four key themes: (1) integrating captions with figures and text, (2) bridging gaps between language proficiency and domain expertise, (3) leveraging multiple AI-generated suggestions, and (4) adapting to diverse writing norms. These findings provide actionable design insights for developing AI writing assistants that better support researchers in creating effective scientific figure captions."
}

@inproceedings{tang-etal-2025-using,
  title = "Using Contextually Aligned Online Reviews to Measure {LLM}s' Performance Disparities Across Language Varieties",
  author = "Tang, Zixin  and
    Huang, Chieh-Yang  and
    Li, Tsung-Chi  and
    Ng, Ho Yin Sam  and
    Huang, Hen-Hsen  and
    Huang, Ting-Hao {`}Kenneth{'}",
  editor = "Chiruzzo, Luis  and
    Ritter, Alan  and
    Wang, Lu",
  booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers)",
  month = apr,
  year = "2025",
  address = "Albuquerque, New Mexico",
  publisher = "Association for Computational Linguistics",
  url = "https://aclanthology.org/2025.naacl-short.29/",
  doi = "10.18653/v1/2025.naacl-short.29",
  pages = "342--355",
  ISBN = "979-8-89176-190-2",
  abstract = "A language can have different varieties. These varieties can affect the performance of natural language processing (NLP) models, including large language models (LLMs), which are often trained on data from widely spoken varieties. This paper introduces a novel and cost-effective approach to benchmark model performance across language varieties. We argue that international online review platforms,such as Booking.com, can serve as effective data sources for constructing datasets that capture comments in different language varieties from similar real-world scenarios, like reviews for the same hotel with the same rating using the same language (e.g., Mandarin Chinese) but different language varieties (e.g., Taiwan Mandarin, Mainland Mandarin). To prove this concept, we constructed a contextually aligned dataset comprising reviews in Taiwan Mandarin and Mainland Mandarin and tested six LLMs in a sentiment analysis task. Our results show that LLMs consistently underperform in Taiwan Mandarin."
}

@inproceedings{ng2025understanding,
  title={Understanding How Paper Writers Use AI-Generated Captions in Figure Caption Writing},
  author={Ng, Ho Yin and Hsu, Ting-Yao and Min, Jiyoo and Kim, Sungchul and Rossi, Ryan A and Yu, Tong and Jung, Hyunggu and Huang, Ting-Hao ‘Kenneth’},
  booktitle={International Workshop on AI for Transportation},
  pages={173--192},
  year={2025},
  organization={Springer}
}

@ARTICLE{10918879,
  author={Weng, Yu-Hsiang and Han, Ping-Hsuan and Chang, Kuan-Ning and Lin, Chi-Yu and Lin, Chia-Hui and Ng, Ho Yin and Chou, Chien-Hsing and Chiu, Wen-Hsin},
  journal={IEEE Transactions on Visualization and Computer Graphics}, 
  title={Hit Around: Substitutional Moving Robot for Immersive and Exertion Interaction with Encountered-Type Haptic}, 
  year={2025},
  volume={31},
  number={5},
  pages={3569-3579},
  abstract={Previous works have shown the potential of immersive technologies to make physical activities a more engaging experience. With encountered-type haptic feedback, users can perceive a more realistic sensation for exertion interaction in substitutional reality. Although substitutional reality has utilized physical environments, props, and devices to provide encountered-type haptic feedback, these cannot withstand the fierce force of humans and do not give feedback when users move around simultaneously, such as in combat sports. In this work, we present Hit Around, a substitutional moving robot for immersive and exertion interaction, in which the user can move and punch the virtual opponent and perceive encountered-type haptic feedback anywhere. We gathered insight into immersive exertion interaction from three exhibitions with iterative prototypes, then designed and implemented the hardware system and application. To understand the ability of mobility and weight loading, we conducted two technical evaluations and a laboratory experiment to validate the feasibility. Finally, a field deployment study explored the limitations and challenges of developing immersive exertion interaction with encountered-type haptics.},
  keywords={Robots;Haptic interfaces;Training;Force;Sports;Virtual environments;Tactile sensors;Prototypes;Legged locomotion;Games;Immersive Exertion Interaction;Substitutional Robot;Encountered-Type Haptic;Virtual Reality;Boxing},
  doi={10.1109/TVCG.2025.3549556},
  ISSN={1941-0506},
  month={May},}

@INPROCEEDINGS{10771086,
  author={Ng, Ho Yin and He, Zeyu and Huang, Ting-Hao ‘Kenneth’},
  booktitle={2024 IEEE Visualization and Visual Analytics (VIS)}, 
  title={What Color Scheme is More Effective in Assisting Readers to Locate Information in a Color-Coded Article?}, 
  year={2024},
  volume={},
  number={},
  pages={291-295},
  abstract={Color coding, a technique assigning specific colors to cluster information types, has proven advantages in aiding human cognitive activities, especially reading and comprehension. The rise of Large Language Models (LLMs) has streamlined document coding, enabling simple automatic text labeling with various schemes. This has the potential to make color-coding more accessible and benefit more users. However, the impact of color choice on information seeking is understudied. We conducted a user study assessing various color schemes’ effectiveness in LLM-coded text documents, standardizing contrast ratios to approximately 5.55:1 across schemes. Participants performed timed information-seeking tasks in color-coded scholarly abstracts. Results showed non-analogous and yellow-inclusive color schemes improved performance, with the latter also being more preferred by participants. These findings can inform better color scheme choices for text annotation. As LLMs advance document coding, we advocate for more research focusing on the "color" aspect of color-coding techniques.},
  keywords={Annotations;Visual analytics;Large language models;Focusing;Color;Encoding;Labeling;Color;Color coding;Information seeking;Text visualization;Document},
  doi={10.1109/VIS55277.2024.00066},
  ISSN={2771-9553},
  month={Oct},}

@inproceedings{10.1145/3550472.3558406,
  author = {Mendez S., Luis Andres and Ng, Ho Yin and Lim, Zin Yin and Lu, Yi-Jie and Han, Ping-Hsuan},
  title = {MovableBag: Substitutional Robot for Enhancing Immersive Boxing Training with Encountered-Type Haptic},
  year = {2022},
  isbn = {9781450394734},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3550472.3558406},
  doi = {10.1145/3550472.3558406},
  abstract = {Boxing is a combat sport that involves fierce movement and close body contact between two boxers. In addition to the boxer, the coach is also essential for training. The coach could be injured or too tired to continue consecutive training sessions for many boxers. Thus, we present MovableBag, a substitutional robot that provides room-scale mobility and the ability to carry sport props with an easier tracking setup to simulate the boxing opponent or coach in virtual reality with encountered-type haptic feedback. In addition, we integrate our system with a wireless VR headset and develop a VR boxing application to show the capacity of the device and inform future studies.},
  booktitle = {SIGGRAPH Asia 2022 XR},
  articleno = {10},
  numpages = {2},
  keywords = {Boxing, Encountered-Type Haptic, Immersive Training, Substitutional Robot, Virtual Reality},
  location = {Daegu, Republic of Korea},
  series = {SA '22}
}

@inproceedings{Ng2022PressySoftiesCSCW,
  title     = {PressySofties: Explore Multi-player Squeeze Interaction with Conductive Fabric Cubes},
  author    = {Ng, Ho Yin and Lin, Chia-Hui and Lim, Zin Yin and Lu, Yi-Jie and Lin, Chu-Yu and Han, Ping-Hsuan},
  booktitle = {ACM Conference on Computer-Supported Cooperative Work and Social Computing 2022, Invited Demos},
  year      = {2022}
}

⚠ auto-generated · verify

@inproceedings{Mendez2022MovableBagMobile,
  title     = {MovableBag: Integrating Haptics and Visual Feedback on Mobile Devices to Enhance the Virtual Reality Experience of Sport Training},
  author    = {Mendez S., Luis Andres and Ng, Ho Yin and Lim, Zin Yin and Lu, Yi-Jie and Han, Ping-Hsuan},
  booktitle = {Proceedings of the 8th Annual Conference of Taiwanese Association of Computer-Human Interaction, Demos},
  address   = {Kaohsiung, Taiwan},
  year      = {2022}
}

⚠ auto-generated · verify

@inproceedings{Ng2022PressySoftiesTAICHI,
  title     = {PressySofties: Utilize Conductive-Cloth Cube to Explore Squeeze Interaction Among Multi-Users},
  author    = {Ng, Ho Yin and Lin, Chia-Hui and Lim, Zin Yin and Lu, Yi-Jie and Lin, Chi-Yu and Han, Ping-Hsuan},
  booktitle = {Proceedings of the 8th Annual Conference of Taiwanese Association of Computer-Human Interaction, Demos},
  address   = {Kaohsiung, Taiwan},
  year      = {2022}
}

⚠ auto-generated · verify

@inproceedings{10.1145/3544793.3561316,
  author = {Mendez, Luis Andres and Ng, Ho Yin and Han, Ping-Hsuan},
  title = {MovableBag: Exploring Asymmetric Interaction for Multi-user Exergame in Extended Reality},
  year = {2023},
  isbn = {9781450394239},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3544793.3561316},
  doi = {10.1145/3544793.3561316},
  abstract = {Previous research shows that virtual reality (VR) exergames motivate users’ physical activity with immersion and enjoyment. However, it limits the movement for players to perform intense exercises in a small space with a complex tracking setup. In addition, asymmetric interaction of multi-user exergame in the same physical space is also lacking in the discussion in extended reality (XR). So, we present MovableBag, a substitutional robot that can be adapted to multi-user exergames with prop. It also serves as an agent to allow users from the physical world to interact with the VR user to become an XR application. Additionally, a preliminary study has been conducted, and a multi-user boxing exergame is designed to investigate the usage of the work. It also informs future design for more use cases in XR for asymmetric interaction between VR and non-VR users.},
  booktitle = {Adjunct Proceedings of the 2022 ACM International Joint Conference on Pervasive and Ubiquitous Computing and the 2022 ACM International Symposium on Wearable Computers},
  pages = {515–519},
  numpages = {5},
  keywords = {Asymmetric Interaction, Encountered-type Haptic, Extended Reality, Multi-user Exergame, Substitutional Robot},
  location = {Cambridge, United Kingdom},
  series = {UbiComp/ISWC '22 Adjunct}
}

@INPROCEEDINGS{9859433,
  author={Chu, Chain Yi and Ng, Ho Yin and Lin, Chia Hui and Han, Ping Hsuan},
  booktitle={2022 IEEE International Conference on Multimedia and Expo Workshops (ICMEW)}, 
  title={PressyCube: An Embeddable Pressure Sensor with Softy Prop for Limb Rehabilitation in Immersive Virtual Reality}, 
  year={2022},
  volume={},
  number={},
  pages={1-1},
  abstract={In this demo, we present PressyCube, an embeddable pressure sensor that allows collecting physical force data from different types of exercises and body parts via Bluetooth for home-based rehabilitation. Currently, most home-based rehabilitation devices are focused on one specific body part only. Thus, we designed the pressure sensor that can be embedded with other props to extend its applications. We also demonstrate our work for fulfilling varying rehabilitation needs in the virtual environment with a head-mounted display by a swimming simulation exergame for players to work out using different exercises.},
  keywords={Pressure sensors;Solid modeling;Head-mounted displays;Bluetooth;Multimedia systems;Conferences;Force;Exergame;Rehabilitation;Virtual Reality;Human-Computer Interaction},
  doi={10.1109/ICMEW56448.2022.9859433},
  ISSN={},
  month={July},}