Terms of Participation.

01 / SCOPE

About these terms

These terms apply to the organizer-maintained FinReason Cup website and participant services operated for the challenge. FinReason Cup is listed as Challenge 03 of the IEEE Big Data Cup 2026. The organizer team is led by The Fin AI, with contributors affiliated with several universities. Those affiliations do not imply institutional sponsorship. These terms do not state that IEEE or an affiliated university authored or legally endorsed them.

By using an organizer-operated participant service or submitting competition materials, a team agrees to these terms and the task-specific rules published with the applicable verified release.

Deadline extension, 18 October 2026: Task 1 and Task 3 Test submissions are due on 25 October 2026, 23:59 Anywhere on Earth (AoE, UTC−12). Final results for Tasks 1 and 3 are scheduled for 27 October. Task 2 also has additional live-trading rounds on 19–23 October, with its final action due on 23 October at 15:25 New York time. Per-round deadlines still apply. Working Notes for all tasks are due separately on 2 November 2026, 23:59 AoE.

For Task 1, required solution materials mean the canonical Test prediction ZIP for all 928 questions, submitted through the Task 1 Test portal. Every team seeking final ranking must submit it; Development uploads alone do not qualify. No separate source-code archive, model weights, repository link, or additional report is required by 25 October. Code and reproducibility materials may be shared voluntarily with Working Notes. Teams seeking final ranking and awards must submit the separate Working Notes paper through CyberChair SC03 by 2 November 2026, 23:59 AoE.

02 / OFFICIAL ROUTES

Use the route assigned to each item

• Submit the challenge paper through the FinReason Cup SC03 track in CyberChair.
• Submit Task 1 competition files only through the verified development or test service linked from the Task 1 participant hub.

Task 1 does not require pre-registration, organizer approval, an access code, or a participant account. Development submission uses Team Name and a canonical ZIP. Test submission uses the same Team Name, a Contact Email, and a canonical ZIP. The Contact Email is not a login, is never published, and is used only for submission identification, submission-related support, matching final results to the related challenge paper, and selecting the latest accepted test submission and handling repeat uploads through a non-public pseudonymous identifier.

03 / PARTICIPATION

Team and submission conduct

Teams must use one consistent Team Name, provide an accurate Contact Email for test submissions, and submit only materials they are permitted to use. Submissions must follow the published schema and must not contain malware, credentials, hidden network calls, or material intended to access private evaluation data or disrupt the service.

Task 3 cases are real SEC filings and carry their own identifiers, so a team can work out which filing a case came from. Deriving the two reported values from the filing materials supplied with the case is the task. Reading either value out of an external source that already publishes data-quality findings for that filing — or out of any other record of the expected answer — is not, and a submission produced that way is not a valid entry, whether or not the source is public. The organizers may ask a team to describe how a result was produced before confirming a final ranking.

Attempts to obtain hidden answers, bypass submission constraints, evade applicable service safeguards, impersonate another team, or interfere with evaluation integrity may be rejected or removed from ranking. The organizers will document any material enforcement decision and give the affected team a contact route.

04 / TASK 3 TEST REVISIONS

Which Task 3 Test submission counts

There is no daily or total Task 3 Test submission-count limit before 25 October 2026, 23:59 AoE. Use the same Team Name and Contact Email for every revision. The latest accepted, valid Test submission before the deadline supersedes earlier submissions for final evaluation. Invalid or rejected uploads do not replace the last accepted submission. An identical repeat upload returns its original receipt and acceptance time and does not create a new latest submission. Earlier records remain private for audit purposes. Test scores and ranks remain hidden until final results are released.

The last submission decides the result rather than the best one. The Test phase returns no score precisely so that the held-out cases cannot be used to choose between a team’s own candidates; ranking the best of several uploads would hand back that choice.

04 / EVALUATION

Task 1 scores and results

Test submission policy update, 16 September 2026: there is no daily or total Test submission-count limit before 25 October 2026, 23:59 AoE. Use the same Team Name and Contact Email for every revision. The latest accepted, valid Test submission before the deadline supersedes earlier submissions for final evaluation. Invalid or rejected uploads do not replace the last accepted submission. An identical repeat upload returns its original receipt and acceptance time and does not create a new latest submission. This policy supersedes previous Test submission-count limits; Development limits are unchanged.

Public training materials may be used as described in their release documentation. Accepted development submissions return Final answer and Reasoning steps scores, a receipt, and current rank, and each team's best eligible result appears on the public development leaderboard. Test uploads receive public-format validation feedback and an acceptance receipt when accepted. They do not receive an online score, rank, answer-correctness feedback, score-derived signal, or scored test leaderboard entry. The organizers perform official test evaluation after submissions close.

Development scores are feedback for the active development phase. Official results are determined from the frozen task contract, eligible submission, and organizer evaluation record. The organizers may correct a demonstrable technical or scoring error and will disclose a material correction that changes published results.

05 / PUBLIC RESULTS

Leaderboard visibility

Each team's best eligible development result appears on the public development leaderboard. Public fields are limited to Team Name, rank, aggregate scores, and acceptance time. Raw submissions, Contact Email, private identifiers, hidden evaluation data, and evaluation receipts are not public leaderboard fields.

06 / PAPER

Challenge paper and publication

Include author names and affiliations in the Working Notes PDF. FinReason follows the conference's single-blind review policy. Working Notes must cite the FinReason Cup overview and the overview for every task entered. Multi-task papers cite the Cup overview once and include all applicable task overviews. Task 1 papers must additionally cite the FinChain benchmark paper (ACL 2026). Task 2 and Task 3 papers must also cite Herculean. Task 3 papers must additionally cite FinAuditing (SIGIR 2026). Copy or download the BibTeX from the citation guidance. The Cup, Task 1, and Task 3 entries include proposed author lists; Task 2 authors await confirmation. Refresh these provisional entries before final paper submission.

Teams seeking final ranking and awards must submit Working Notes (a challenge paper) of up to 10 pages total, including references, in the IEEE two-column conference format through CyberChair SC03 by 2 November 2026, 23:59 Anywhere on Earth. The organizers review the papers and decide acceptance, with feedback and decisions scheduled for 9 November. Accepted teams must submit camera-ready papers by 16 November 2026, 23:59 Anywhere on Earth.

Selected accepted papers may be included in the conference proceedings, subject to conference publication, camera-ready, registration, and presentation requirements. Submission, ranking, or a certificate does not guarantee paper acceptance or publication. The organizer overview is due to the conference on 20 November 2026; the award announcement is scheduled for 25 November 2026.

07 / RIGHTS

Resources and participant materials

Each released dataset, code package, or other resource is governed only by the license published with that resource. These terms do not create a license for a resource that does not include one.

Participants retain their rights in their papers, code, models, and submission files. Participants permit the organizers to receive, validate, securely store, reproduce, and evaluate submitted materials as needed to operate the challenge, investigate integrity issues, and produce results. The organizers will not publish raw submission files or participant source code unless the applicable rules state that publication is required or the team separately authorizes it.

08 / CERTIFICATES AND PRIZES

Certificates and prizes

FinReason does not offer cash prizes. Registration support is not confirmed at this time, and the organizers do not promise registration funding. A team will receive a participation certificate if it submits both (1) at least one valid final solution under the applicable task rules and deadline (25 October 2026, 23:59 AoE for Tasks 1 and 3; the published live-trading deadlines for Task 2) and (2) Working Notes through CyberChair SC03 by 2 November 2026, 23:59 Anywhere on Earth. Winning teams will receive a winner certificate.

Additional award categories remain provisional until published. Certificates do not imply challenge-paper acceptance or publication.

09 / SERVICES

External platforms and rule updates

CyberChair, GitHub, Hugging Face, the conference website, and other external services may apply their own terms and privacy notices. The organizers may update technical rules, dates, or service routes when needed for a secure and fair competition. Material changes will be dated and published on the organizer-maintained website before they apply whenever practical.