AIdea

In Progress

day

min

sec

Time to team-up deadline

0 / 0

participants / team

Topic provider

Graduate Institute of Information Management, National Taipei University, NTPU

The Graduate Institute of Information Management, National Taipei University (NTPU) is dedicated to cultivating professionals in the fields of information management and digital innovation. Its curriculum centers on information technology, equipping students with advanced skills in artificial intelligence, big data analytics, cloud computing, mixed reality, fintech, and blockchain, while also emphasizing key areas such as information security compliance and digital sustainability. Through interdisciplinary practical applications, the program extends into digital marketing, digital transformation, consumer behavior insights, and sports and leisure management, ensuring students can translate technical expertise into tangible industry value.
Academically, the Institute emphasizes a dual focus on research and practical application. Close industry-academia collaborations and cross-disciplinary integration provide students with substantial hands-on experience. The program fosters the ability to explore emerging information technologies, develop system applications, and plan digital marketing strategies, while also cultivating rigorous academic writing and independent research skills. Students joining the Institute will grow into adaptable professionals capable of making a significant impact in information management, smart services, and emerging technology applications.

Introduction

As the United Nations Sustainable Development Goals (SDGs) and global net-zero policies accelerate, ESG disclosure has become a defining measure of corporate credibility. Yet many sustainability reports still contain vague or unverified claims, highlighting an urgent need for intelligent verification tools—especially for Traditional Chinese ESG texts, where language resources remain limited.

The “ESG Sustainability Commitment Verification Competition 2026” invites you to tackle this real-world challenge. Using the VeriPromiseESG4K Annotated Corpus, built from authentic industry scenarios, participants will develop NLP models that identify sustainability commitments and assess their supporting evidence.

This is an opportunity to apply AI to a high-impact global issue, sharpen your skills in ESG analytics and model development, and contribute to greater transparency and trust in corporate sustainability disclosures.

This competition provides 4,000 annotated Traditional Chinese ESG entries drawn from real corporate sustainability reports, forming the VeriPromiseESG4K dataset. Participants will apply AI technology such as natural language processing and large language models to develop NLP models capable of automatically identifying, analyzing, and verifying corporate sustainability commitments.

Participating teams will utilize the “VeriPromiseESG4K Annotated Corpus” to develop Natural Language Processing (NLP) models addressing the following four core tasks and generate prediction results for the test dataset:

Subtask 1: Commitment Classification
Determine whether the given text expresses a concrete corporate commitment toward future actions.
Subtask 2: Timeline Classification
Based on semantic inference, determine the expected completion timeframe of the identified commitment (calculated from the publication year 2024 of the ESG report).
Subtask 3: Evidence Identification
Determine whether the identified commitment sentence is supported by action plans or existing records (e.g., data, methodologies, measures, or any concrete implementation details).
Subtask 4: Clarity Classification
If evidence is present, evaluate whether the evidence statement is semantically explicit; easy to understand; free from vague wording (e.g., “continuously advancing,” “striving to improve”).

Eligibility

Student Division: Open to currently enrolled students with official student status in the Republic of China (including graduate and senior high school students).
General Public Division: If any team member does not meet the eligibility requirements of the Student Division (including students currently enrolled in institutions outside the Republic of China), the entire team will be classified under the Open Division.

Registration

Each team member must register through the AI CUP Registration System(first-time users may refer to the AI CUP Registration System guide for instructions).Participants must follow the on-screen registration procedures, complete all required information, and finalize team formation to complete competition enrollment.
Team names will be automatically assigned by the system; participants are not permitted to choose their own team names.
Within 1–2 business days after completing competition enrollment via the AI CUP Registration System,an account activation notice and password for theAIdea Platformwill be sent to the participant’s registered email address. After completing the activation steps indicated in the email, all team members may access theAIdea Platform to participate in the competition.
If participants are registering as part of a course requirement, they should obtain the course code from the course instructor or teaching assistant and enter it during registration. This will enable the instructor to access competition results for course evaluation purposes.

For inquiries, please contact: yyteng@mail.ntpu.edu.tw

Prize

In this Competition, the top 15 teams will have the opportunity to share a total prize pool of NT$250,000.

Student Division: Valid proof of enrollment (e.g., student ID copy or official enrollment certificate) is required for award eligibility.

Award	Quota	Prize for Each Team
First Prize	1	NT$80,000 + A printed certificate from the Ministry of Education (& an electronic certificate issued by the Project Office)
Second Prize	1	NT$50,000 + A printed certificate from the Ministry of Education (& an electronic certificate issued by the Project Office)
Third Prize	1	NT$30,000 + A printed certificate from the Ministry of Education (& an electronic certificate issued by the Project Office)
Excellence Award	2	NT$10,000 + A printed certificate from the Ministry of Education
Honorable Mention Award	10	NT$7,000 + A printed certificate from the Ministry of Education

General Public Division

Award	Quota	Prize for Each Team
First Prize	1	Electronic certificate issued by the Ministry of Education AI Competition Project Office
Second Prize	1	Electronic certificate issued by the Ministry of Education AI Competition Project Office
Third Prize	1	Electronic certificate issued by the Ministry of Education AI Competition Project Office
Excellence Award	2	Electronic certificate issued by the Ministry of Education AI Competition Project Office
Honorable Mention Award	10	Electronic certificate issued by the Ministry of Education AI Competition Project Office

*Note: The organizer reserves the right to adjust the number of awards based on submission quantity and quality. Awards may be withheld if entries do not meet the required standards.

Activity time

The competition will begin on Wednesday, March 4, 2026 (Taiwan Time, UTC+8) and will officially conclude with the announcement of results on Thursday, July 23, 2026. The detailed schedule is as follows:

Date	Item
2026/03/04(Wed)-2026/04/28(Tue)	Registration Opens & First Portion of the Training Set Release
March 2026 (the exact detail will be announced on the official competition website)	Regional Hands-on Workshops: Northern Region: University of Taipei (UTaipei); Central Region: Providence University (PU); Southern Region: National Kaohsiung University of Science and Technology (NKUST).
2026/04/28(Tue)-2026/06/03(Wed)	Second Portion of the Training Set Release
2026/06/03(Wed)-2026/06/10(Wed)	Validation Set Release
2026/06/10(Wed)-2026/06/17(Wed)	Test Set Release & Prediction Submission
2026/06/23(Tue)	Preliminary Results Announcement
2026/06/24(Wed)-2026/06/30(Tue)	Submission of Additional Deliverables (Report and Code)
2026/07/01(Wed)-2026/07/14(Tue)	Evaluation Period
2026/07/23(Thu)	Final Ranking Announcement
2027/03	Award Ceremony

Evaluation Criteria

Participants in VeriPromiseESG 2026 (hereinafter referred to as “the competition”) are required to develop an AI model capable of completing four core tasks, based on three datasets provided by the organizer: the Training Data, Validation Data, and Test Data, along with annotated sample data. The stages of the Competition are as follows:

Stage 1: The organizer releases annotated sample data and opens registration.
Stage 2: Participating teams train their models using the Training Data and generate predictions on the Validation Data, uploading results to the online Public Leaderboard.
Stage 3: After the organizer releases the full Private Dataset, teams must upload their predictions for the Test Data to the platform before the deadline. Each team may submit up to three times per day.
Stage 4: At the close of the Competition, final rankings will be determined based on results from the Private Dataset and announced on the Private Leaderboard.

Final rankings will be calculated using a weighted composite score across the four tasks to evaluate overall system performance (see “Task Evaluation Criteria” and “Evaluation Formula” for details).
All teams must submit the required technical report and original source code within the specified timeline to verify the absence of manual adjustments, misconduct, or plagiarism. Teams that fail to submit the required materials on time will not be included in the final ranking.

Task Evaluation Criteria

The judging panel, appointed by the organizer and composed of industry professionals and academic experts, will calculate the final score using a weighted average of the four subtasks:

Evaluation Criteria	Description	Weight
Commitment Classification	Balance between precision and recall in identifying ESG commitment statements	20%
Evidence Identification	Ability to determine whether commitments are sufficiently supported by evidence	30%
Clarity Classification	Three-class classification performance for evaluating evidence quality	35%
Timeline Classification	Four-class classification performance for predicting appropriate verification timing	15%

Evaluation Formula

$$Total Score=(\textit{Commitment Classification F1 Score}×0.20)+(\textit{Evidence Identification F1 Score}×0.30)+(\textit{Clarity Classification F1 Score}×0.35)+(\textit{Timeline Classification F1 Score}×0.15)$$

Rules

All teams must carefully read the following provisions. In the event of disputes regarding rights or violations, the organizer reserves the right to revoke participation or award eligibility. Teams shall bear full responsibility for any consequences. If prizes have already been awarded, the organizer reserves the right to reclaim them.

The total number of daily submissions for the Public Dataset and Private Dataset predictions is limited to 3 per day. The system will display the highest-scoring submission on the Leaderboard.
The organizer reserves the right to adjust the dataset during the competition if necessary.
Without prior notice, the organizer may disqualify any team under the following circumstances:

Verified evidence of plagiarism, cheating, or fraud;
Infringement of intellectual property rights;
Attacks on the leaderboard system;
Actions affecting other teams and compromising fairness;
Violations of competition regulations.

All reports, code, and prediction results must be submitted before the deadline. Late submissions, revisions, or supplementary submissions will not be accepted. Only teams that complete the submission within the deadline will be ranked. Failed or incomplete submissions will be considered a withdrawal.
Manual annotation or modification of the released test dataset is strictly prohibited. All predictions must be automatically generated by programs to ensure fairness.
Teams may not privately share code or features, but may engage in discussion within the official forum.
Teams must respect all judging decisions and final results.
All intellectual property rights related to submitted materials remain with the teams; however, teams grant the organizer a royalty-free license for promotional, archival, and exhibition purposes.
By completing registration, teams agree to comply with all competition rules and regulations.
The organizer reserves the right to interpret, amend, suspend, or terminate the competition rules if necessary.