IEEE BIG DATA CUP · 2026
Agentic financial reasoning,
hedging, and audit verification.
Three tasks for building and evaluating financial AI systems. Explore the data, develop your approach, and submit your results.
14–17 December 2026 Phoenix, Arizona
01 / COMPETITION TASKS
Each participant hub contains its data, rules,
and current submission options.
Solve multi-step financial problems with a typed final answer and values for the published reasoning steps.
2,900 labeled training cases, 290 labeled local-development cases, 580 unlabeled leaderboard-development questions, and 928 public test questions.
Manage paired asset positions with zero net dollar exposure using point-in-time prices, news, and corporate filings.
HERCULEAN training data is available. Evaluation windows, execution costs, and final submission rules will follow with the scorer. The displayed measures are provisional until final scoring rules are published.
Verify reported numeric facts against their calculation context in organizer-packaged SEC EDGAR XBRL filings.
Explore the starter kit, validator, and public practice workflow. See the task hub for phase availability and held-out evaluation updates. This task evaluates targeted numeric checks, not a full financial-statement audit.
02 / EVALUATION
Use the published data, schemas, and validators. Task 1 development uploads return separate final-answer and reasoning-step scores.
Task 1 allows unlimited submissions before the deadline. Keep the same Team Name and Contact Email; the latest accepted, valid version is used.
Task 1 Test scores and ranks are now published under the frozen evaluation configuration, with a known-question inconsistency noted on the results page. Test answers remain private. Follow each hub for task-specific evaluation rules.
03 / KEY DATES
All dates are in 2026.
AoE = Anywhere on Earth (UTC−12).
Final system submissions and required solution materials are due. Task 2 live-trading actions must meet their earlier per-round deadlines on 19–23 October.
Verified final results are scheduled for 27 October. Task 2 interim performance updates follow its live-trading settlements.
Submit the team’s challenge paper through CyberChair SC03, up to 10 pages total including references.
The organizers return paper feedback and acceptance decisions.
Accepted teams submit their revised, camera-ready papers.
The organizers submit the challenge overview to the conference.
Award announcement. Verified final results are scheduled for 27 October.
Phoenix, Arizona.
The Task 1 and Task 3 Test cutoff is 25 October, 23:59 AoE (26 October, 11:59 UTC). Task 2 adds 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 submission is open through CyberChair SC03 until 2 November 2026, 23:59 AoE.
04 / WORKING NOTES
Prediction files go through the task hubs.
Working Notes go through CyberChair SC03.
Teams seeking final ranking and awards must submit a challenge paper through the official FinReason Cup track in CyberChair.
Working Notes must cite the Cup overview and the overview for every task you enter. View required references and BibTeX.
The schedule was extended on 18 October 2026. Task 1 and Task 3 Test submissions close on 25 October 2026, 23:59 AoE. Task 2 also has additional live-trading rounds on 19–23 October; its final action is due on 23 October at 15:25 New York time. Per-round deadlines still apply.
The organizers decide paper acceptance; selected accepted papers may enter the proceedings subject to conference requirements. Publication is not guaranteed.
Task 1 data and tools are released through the participant hub. No pre-registration or access code is required.
Development: accepts Team Name plus a 580-row predictions ZIP, immediately returns Final answer, Reasoning steps, a receipt, and current rank, and publishes the team’s best eligible result on the public leaderboard.
Test: accepts Team Name, Contact Email, and a 928-row predictions ZIP, then shows format feedback and an acceptance receipt with no online score or rank.
05 / REFERENCES
Use the supplied BibTeX
to keep reference details consistent.
Your Working Notes must cite the FinReason Cup overview and the overview for each task you participate in. If you enter multiple tasks, cite the Cup overview once and include every relevant task overview.
These organizer-supplied BibTeX entries are provisional. The manuscripts are in preparation; The Cup and Task 1 entries include proposed author names and order. The Task 2 author list still uses Pending until its task lead confirms it. Publication details are not yet assigned. Use these citation keys now and refresh the entries here before submitting your final paper.
@unpublished{finreason2026overview,
author = {Xie, Zhuohan and Wang, Yan and Jiang, Yuechen and Peng, Xueqing and Qian, Lingfei and Cao, Yupeng and Chen, Yankai and Liu, Xue (Steve) and Nakov, Preslav},
title = {Overview of the {FinReason Cup 2026}: Agentic Financial Reasoning, Hedging, and Audit Verification},
year = {2026},
note = {Manuscript in preparation; proposed author list and order},
url = {https://the-finai.github.io/IEEE-bigdata-cup/}
}@unpublished{finreason2026task1overview,
author = {Xie, Zhuohan and Wang, Yan and Jiang, Yuechen and Peng, Xueqing and Qian, Lingfei and Cao, Yupeng and Chen, Yankai and Liu, Xue (Steve) and Nakov, Preslav},
title = {Overview of the {FinReason Cup 2026} Task 1: Verifiable Financial Chain Reasoning},
year = {2026},
note = {Manuscript in preparation; proposed author list and order},
url = {https://the-finai.github.io/IEEE-bigdata-cup/task1/}
}@inproceedings{finreason2026task2overview,
author = {{Pending}},
title = {Overview of the {FinReason Cup 2026} Task 2:
Market-Neutral Hedging},
booktitle = {2026 IEEE International Conference on Big Data (BigData)},
year = {2026},
address = {Phoenix, Arizona, USA},
note = {Overview manuscript in preparation; author list pending},
url = {https://the-finai.github.io/IEEE-bigdata-cup/task2/}
}@inproceedings{finreason2026task3overview,
author = {Wang, Yan and Xie, Zhuohan and Jiang, Yuechen and Qian, Lingfei and Peng, Xueqing and Cao, Yupeng and Chen, Yankai and Nakov, Preslav and Liu, Xue (Steve)},
title = {Overview of the {FinReason Cup 2026} Task 3:
Financial Audit Verification},
booktitle = {2026 IEEE International Conference on Big Data (BigData)},
year = {2026},
address = {Phoenix, Arizona, USA},
note = {Manuscript in preparation; proposed author list and order},
url = {https://the-finai.github.io/IEEE-bigdata-cup/task3/}
}Zhuohan Xie et al. (2026) · ACL 2026, Volume 1: Long Papers, pages 14529–14553.
Task 1 Working Notes must cite the benchmark paper. Copy the verified BibTeX below into your bibliography. It uses the published ACL 2026 reference from the ACL Anthology.
@inproceedings{xie-etal-2026-finchain,
title = "{F}in{C}hain: A Symbolic Benchmark for Verifiable Chain-of-Thought Financial Reasoning",
author = "Xie, Zhuohan and
Orel, Daniil and
Thareja, Rushil and
Sahnan, Dhruv and
Madmoun, Hachem and
Zhang, Fan and
Banerjee, Debopriyo and
Georgiev, Georgi Nenkov and
Peng, Xueqing and
Qian, Lingfei and
Huang, Jimin and
Su, Jinyan and
Singh, Aaryamonvikram and
Xing, Rui and
Elbadry, Rania and
Xu, Chen and
Li, Haonan and
Koto, Fajri and
Koychev, Ivan and
Chakraborty, Tanmoy and
Wang, Yuxia and
Lahlou, Salem and
Stoyanov, Veselin and
Ananiadou, Sophia and
Nakov, Preslav",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.662/",
doi = "10.18653/v1/2026.acl-long.662",
pages = "14529--14553",
ISBN = "979-8-89176-390-6"
}Xueqing Peng et al. (2026) · arXiv:2605.14355.
Task 2 and Task 3 Working Notes must cite Herculean in addition to the Cup overview and their task overview.
@misc{peng2026herculeanagenticbenchmarkfinancial,
title = {{Herculean}: An Agentic Benchmark for Financial Intelligence},
author = {Xueqing Peng and
Zhuohan Xie and
Yupeng Cao and
Haohang Li and
Lingfei Qian and
Yan Wang and
Vincent Jim Zhang and
Huan He and
Xuguang Ai and
Linhai Ma and
Ruoyu Xiang and
Yueru He and
Yi Han and
Shuyao Wang and
Yuqing Guo and
Mingyang Jiang and
Yilun Zhao and
Youzhong Dong and
Xiaoyu Wang and
Yankai Chen and
Ye Yuan and
Qiyuan Zhang and
Fuyuan Lyu and
Haolun Wu and
Yonghan Yang and
Zichen Zhao and
Yuyang Dai and
Fan Zhang and
Rania Elbadry and
Ayesha Gull and
Muhammad Usman Safder and
Nuo Chen and
Fengbin Zhu and
Tianshi Cai and
Zimu Wang and
Polydoros Giannouris and
Yuechen Jiang and
Zhiwei Liu and
Mohsinul Kabir and
Yuyan Wang and
Yixiang Zheng and
Yangyang Yu and
Weijin Liu and
Wenbo Cao and
Anke Xu and
Peng Lu and
Jerry Huang and
Mingquan Lin and
Prayag Tiwari and
Yijia Zhao and
Víctor Gutiérrez-Basulto and
Xiao-Yang Liu and
Kaleb E Smith and
Jiahuan Pei and
Arman Cohan and
Jimin Huang and
Yuehua Tang and
Alejandro Lopez-Lira and
Xi Chen and
Xue Liu and
Junichi Tsujii and
Jian-Yun Nie and
Sophia Ananiadou},
year = {2026},
eprint = {2605.14355},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2605.14355}
}Yan Wang et al. (2026) · SIGIR 2026, pages 3456–3463.
Task 3 Working Notes must also cite FinAuditing. This BibTeX uses the published SIGIR 2026 version, with the complete published author list and DOI.
@inproceedings{wang2026finauditing,
title = {{FinAuditing}: A Financial Taxonomy-Structured Multi-Document Benchmark for Evaluating {LLMs}},
author = {Yan Wang and
Keyi Wang and
Shanshan Yang and
Jaisal Patel and
Jeff Zhao and
Fengran Mo and
Xueqing Peng and
Lingfei Qian and
Yankai Chen and
Víctor Gutiérrez-Basulto and
Jimin Huang and
Guojun Xiong and
Xiao-Yang Liu and
Jian-Yun Nie},
booktitle = {Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval},
year = {2026},
pages = {3456--3463},
publisher = {Association for Computing Machinery},
address = {Melbourne, Australia},
doi = {10.1145/3805712.3808578},
url = {https://doi.org/10.1145/3805712.3808578}
}Cite the task data, tools, and any other resources used in your system as appropriate.
06 / PARTICIPANT FAQ
Yes. Test submissions are unlimited before 25 October 2026, 23:59 Anywhere on Earth, with no daily or total count limit. Keep the same Team Name and Contact Email. Your latest accepted, valid submission replaces earlier versions for final evaluation; invalid or rejected uploads do not. An identical repeat upload returns its original receipt and acceptance time, rather than creating a new revision. This policy was updated on 16 September 2026.
Every Task 1 team seeking final ranking must submit the canonical ZIP containing predictions.jsonl for all 928 Test questions through the Task 1 Test portal by 25 October 2026, 23:59 AoE. Development uploads alone do not qualify. No separate source-code archive, model weights, repository link, or additional report is required by that deadline. Code and reproducibility materials may be shared voluntarily with Working Notes. Teams seeking final ranking and awards must also submit the separate Working Notes paper through CyberChair SC03 by 2 November 2026, 23:59 AoE.
Keep author names and affiliations in your Working Notes PDF. FinReason follows the conference's published single-blind review policy: reviewers can see author identities. Submit through CyberChair SC03 using the IEEE two-column format, up to 10 pages including references.
Use the official FinReason Cup SC03 track in CyberChair. 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 by 2 November 2026, 23:59 Anywhere on Earth.
No. Task 1 and Task 3 Test submissions are due through their task-specific competition routes by 25 October 2026, 23:59 Anywhere on Earth. Task 2 is also extended: additional live-trading rounds run on 19–23 October, with the final action due on 23 October at 15:25 New York time. Per-round deadlines still apply. Working Notes use CyberChair SC03 and are due separately on 2 November 2026, 23:59 Anywhere on Earth. Use the participant hubs for current task availability and upload instructions.
Working Notes submission is open through CyberChair SC03, with a deadline of 2 November 2026, 23:59 Anywhere on Earth. Papers may be up to 10 pages including references. The competition submission deadline is 25 October, 23:59 AoE. Task 2 adds live-trading rounds on 19–23 October with a final action deadline of 23 October, 15:25 New York time; per-round deadlines still apply. Camera-ready papers are due on 16 November, 23:59 AoE. These dates replace the previously published combined 15 November cutoff.
No. The organizers review the Working Notes and decide paper acceptance, with feedback and decisions scheduled for 9 November. Selected accepted papers may be included in the conference proceedings, subject to conference publication, camera-ready, registration, and presentation requirements. Participation, ranking, or a certificate does not guarantee paper acceptance or proceedings publication.
Task 1 is open for direct participation now. Multi-task participation and the separate Task 2 and Task 3 submission rules will be published with those starter kits.
The Task 1 participant hub provides the training, development, and test downloads. Teams upload predictions directly through separate development and test submission pages. Development returns scores and updates the public leaderboard immediately; test shows format feedback and an acceptance receipt, with no online score or rank. The paper route remains separate through CyberChair SC03.
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 paper acceptance or publication.
No. Choose a consistent team name and submit directly. Development asks for Team Name and a canonical ZIP. Test asks for the same Team Name, a Contact Email, and a canonical ZIP. The Contact Email is private, is not used as a login, 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.
An accepted development upload immediately shows the Final answer score, Reasoning steps score, receipt ID, and current rank, and updates the public development leaderboard. An accepted test upload shows only a receipt; test scores, ranks, diagnostics, and score-derived signals remain hidden until final results.
Email zhuohan.xie@mbzuai.ac.ae for participant support, submission questions, team-name corrections, or privacy requests. Include the team name and task number when applicable, but do not send a submission archive by email.
07 / ORGANIZERS
