Vietnamese Physics Reasoning and Answering Challenge 2026

Important dates

September 18: Training Data release

September 25: Public Test

October 9: Private Test

October 15: Result announcement

October 25: Paper submission

November 5: Acceptance notification

November 12: Camera-ready (= > Proceedings)

November 15: Workshop date

General Description

ViPRA 2026 is a shared task under VLSP 2026 focusing on Vietnamese multimodal understanding and reasoning in the physics education domain. Given a Vietnamese physics question together with associated visual information, such as diagrams, figures, or illustrations, participants are required to identify and generate the correct answer. The task evaluates the ability of AI systems to jointly understand Vietnamese language, interpret visual information, and apply relevant physics knowledge to solve problems. ViPRA aims to establish a challenging benchmark for Vietnamese multimodal question answering and reasoning, and to encourage the development of multimodal large language models and other vision-language approaches for educational applications.

Dataset

The shared task will be based on ViPPS (Vietnamese Physics Problem Solving), a Vietnamese physics question-answering dataset developed for research on multimodal reasoning in physics education. ViPPS contains 5,520 physics questions, including 3,750 questions associated with visual information and 1,770 text-only questions. The visual subset contains 448 unique images, including physics diagrams and illustrations covering topics such as electrical circuits and mechanics.

The primary dataset for ViPRA 2026 will consist of the 3,750 multimodal questions. Each instance contains a Vietnamese physics question and its associated visual context. The dataset will be divided into training, development, and test sets. Ground-truth answers for the test set will be withheld and used for official evaluation.

The task is designed to assess multimodal Vietnamese language understanding and physics reasoning. Many questions require participants to integrate information from the Vietnamese question and visual representation and apply appropriate physics knowledge to obtain the answer. The 1,770 text-only questions may be provided as auxiliary training data or used in an additional track, depending on the final competition design.

Evaluation Metrics

[]

Submission Format

[]

Contact

Zalo Group: []

Registration

https://forms.gle/nG72dFo8a3UVn1DW8

Organizers

• Assoc. Prof. Dr. Quan Thanh Tho, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), qttho@hcmut.edu.vn
• Ms. Le Thi Xinh, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), ltxinh.sdh251@hcmut.edu.vn
• Mr. Nguyen Song Thien Long, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), long.nguyencse2023@hcmut.edu.vn
• Ms. Vo Thi Nhu Quynh, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), quynh.vo01@hcmut.edu.vn
• Mr. Nguyen Huu Nam Phong, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), phong.nguyenhuunam@hcmut.edu.vn
• Ms. Tran Huynh Mai Thao, Faculty of Computer Science and Engineering, Ho Chi Minh City University of Technology (HCMUT), Vietnam National University Ho Chi Minh City (VNU-HCM), thao.tran101205@hcmut.edu.vn

References

1. Example