In collaborative with the Boston Symmetry Group, TAG in DS is proud to announce that the 2nd annual TAG-DS conference,
The Boston TAG Party -- "No AI Without Mathematical Representation!"
to be held at Northeastern University in Boston, MA, USA from August 18-20, 2026.
Follow our twitter account (@TAGinDS) for real-time conference updates and news.
If you have questions, please contact the organizers at info@tagds.com
TAG in DS and the Boston Symmetry Group are proud to present The Boston TAG Party 2026 – a collaborative conference rallying together to ensure, “No AI without Mathematical Representation!”
Much of the data that is fueling current rapid advances in machine learning is high dimensional, structurally complex, and strongly nonlinear. This poses challenges for researcher intuition when they ask (i) how and why current algorithms work and (ii) what tools will lead to the next big break- though. Mathematicians working in topology, algebra, and geometry have more than a century’s worth of finely-developed machinery whose purpose is to give structure to, help build intuition about, and generally better understand spaces and structures beyond those that we can naturally understand.
Building on the success of past TAG-DS workshops, this event will showcase work which brings methods from topology, algebra, and geometry and uses them to help answer challenging questions in machine learning and artificial intelligence. Symmetry as a unifying theme of the three TAG branches, and in collaboration with the Boston Symmetry Group, the 2026 Boston Symmetry Day will form the first day of the event.
Submissions are welcome for both a full archival paper track and a non-archival extended abstract track; only accepted full, archival-track papers will be published in the associated volume. Want to get an idea out there or looking for collaborators on a problem of interest? New this year, we are adding an open-problem/conjecture track!
Please join us for 3 days filled with collaboration and networking activities, exciting talks, illuminating panels, and mathematical ideation to advance and understand machine learning!
June 12 , 2026 - (AOE) June 15, 2026 - (AOE)
(Paper submission deadline extended)
July 15, 2026 - (AOE)
August 3, 2026 - (AOE)
August 12, 2026 - (AOE)
August 18, 2026 - (AOE)
Coming soon!
TAG-DS is offering 3 submission tracks this year:
Full archival paper track (8-pages excluding references, appendices, and supplementary material)
Full archival papers must include novel work, results, and theory consistent with requirements for publication as completed research with clear conclusions. Submissions must stand-alone, independent of appendices and supplementary material.
Accepted full archival papers will be published in the associated PMLR volume
Non-archival extended abstract track (4-pages excluding references)
Non-archival extended abstracts may include work in progress or summaries of work that has been or will be published in alternative venues
Will not be published in the associated PMLR volume
Open-problem/conjecture track (3-pages excluding references)
Open-problem/conjecture submissions should describe important open problems within the TAG space
What are the major questions at the interface of AI and math? We are specifically looking for questions that are broad enough to be accessible and of interest to a significant fraction of attendees but remain focused on TAG topics. Examples might include: a better understanding of how specific mathematical perspectives might apply to machine learning; a theoretical explanation of observed phenomena (e.g., “why do equivariant architectures always…”); or methods of measuring TAG-relevant quantities “how can we measure X in large foundation models?” This track is also appropriate for posing open conjectures to engage the community and seed future collaborations.
Will not be published in the associated PMLR volume
**** Only accepted, full archival-track papers will be published in the associated PMLR volume ****
Mathematically-constrained representation learning
Symmetry in data and learning
Novel architectures
Alternative learning objectives
Training schemes
Robustness
AI for Math
Model Evaluation
Datasets, Explainability
Reduced-order and Energy Efficient Models
Domain-driven Data Analytics
Survey papers relevant to the TAG-DS community
The OpenReview submission website is the same for all three tracks:
Within the submission form you must choose the track to which you wish to submit.
**** Only accepted, full archival-track papers will be published in the associated PMLR volume ****
Please format all submitted papers using the TAG-DS 2026 template:
LaTeX template for paper submissions
The template provides three formatting options:
submission --> default; for all submitted papers prior to acceptance
proceedings --> use for accepted full archival (proceedings) papers
nonproceedings --> use for accepted non-archival extended abstracts and open-problem/conjecture track papers
All submissions should be formatted using the submission option prior to acceptance.
Please contact us at info@tagds.com if you run into any issues with the template.
The TAG-DS LaTeX template is based on the Journal of Machine Learning Research (JMLR) official style files.
In collaboration with TopoBench and GraphUniverse, TAG-DS 2026 is proud to once again host the Topological Deep Learning (TDL) Challenge. The challenge is organized by Guillermo Bernárdez, Lev Telyatnikov, Mathilde Papillon, Marco Montagna, Louisa Cornelis, Louis Van Langendonck, Olga Fink, and Nina Miolane.
The theme of this year's challenge is, "Topological Deep Learning Challenge 2026: Bridging the Gap," with the goal of connecting Topological Neural Networks (TNNs) and Graph Neural Networks (GNNs). For the first time, the TDL Challenge will go beyond implementation to feature a rigorous performance analysis of the submitted models. Through a shared benchmarking ecosystem of GNNs and TNNs, the 2026 TDL Challenge aims to formulate data-driven answers to long-standing scientific questions:
Structural Sensitivity: How do specific graph properties (e.g., severe heterophily) impact the performance of classical GNNs versus their higher-order topological counterparts
The Topological Component: Under what specific data regimes and controlled environments do TDL models consistently provide unique capabilities over standard state-of-the-art GNN approaches (if any)?
Submissions to the TDL challenge are made directly through the challenge website:
https://geometric-intelligence.github.io/topobench/tdl-challenge-2026/index.html
All submissions are due by August 12th, 2026 - (AoE).
***Every submission that meets the requirements will be included in a white paper summarizing the challenge’s findings (planned via PMLR through Topology, Algebra, and Geometry in Machine Learning/Data Science 2026). Authors of qualifying submissions will be offered co-authorship.***
Two winning teams (one per track) will be announced at TAG-DS 2026 during the Awards Ceremony, and will receive the following prizes:
Track 1 (GNNs): 1st place $1,000 USD, 2nd place $400 USD (sponsored by New Theory).
Track 2 (TNNs): 1st place $1,000 USD, 2nd place $400 USD (sponsored by Arlequin AI).
Honorable mentions: $700 USD split across other outstanding submissions (additional evaluation notebook with further benchmarking, particularly challenging implementations, participants who submit multiple high-quality submissions, etc).
Additionally, the Geometric Intelligence Lab, University of California, Santa Barbara and the Intelligent Maintenance and Operations Systems (IMOS) Lab at EPFL in Lausanne, Switzerland are each offering research internships to qualifying teams.
Check out the challenge website for more details and official rules.
Coming soon!
Coming soon!
Curry Student Center
CSC Ballroom (2nd Floor)
Northeastern University
360 Huntington Ave
Boston, MA 02115
To travel from Boston to Northeastern University using public transportation (MBTA):
Subway (T)
Take the Green Line E branch (also known as the “E Line”) outbound towards Heath Street.
Exit the train at the “Northeastern University” station.
Northeastern University is within walking distance from the station.
Commuter Rail
If you are coming from a location outside of Boston and closer to a Commuter Rail station, you can take a Commuter Rail train that services the Ruggles station.
From Ruggles station, it's just a short walk to Northeastern University.
Bus
Various MBTA buses service the Northeastern University area.
You can check the MBTA website or use the Transit app for specific bus routes and schedules, as they may change from time to time.
For the most up to date information on public transportation, please refer to the official Massachusetts Bay Transportation Authority (MBTA) website or Transit app.
To travel to Northeaster University by car, the following (paid) parking garages are close to the conference venue:
Gainsborough Garage
10 Gainsborough Street
Boston, MA 02115
Renaissance Park Garage
835 Columbus Avenue
Boston, MA 02120
Eddie Berman
Northeastern University
Guillermo Bernárdez
University of California
Santa Barbara
Samantha Chen
Oberlin College
Alex Cloninger
University of California
San Diego
Timothy Doster
Pacific Northwest
National Laboratory
Tegan Emerson
Pacific Northwest
National Laboratory,
University of Texas, El Paso
J. Elisenda Grigsby
Boston College
Henry Kvinge
Pacific Northwest
National Laboratory,
University of Washington
Hannah Lawrence
Massachusetts Institute of Technology
Tim Marrinan
Pacific Northwest
National Laboratory
Audun Myers
Pacific Northwest
National Laboratory
Mathilde Papillon
University of California
Santa Barbara
Behrooz Tahmasebi
Harvard University
Lev Telyatnikov
Robin Walters
Northeastern University
Melanie Weber
Harvard University
YuQing Xie
Massachusetts Institute of Technology
Eric Yeats
Pacific Northwest
National Laboratory
The Boston TAG Party is possible in part due to the support of our sponsors. Check out their websites to see what kind of cool things they are working on and meet with them in-person at the conference for more information.
If you, or your organization, are interesting in joining our TAG-DS research community and sponsoring the Boston TAG Party, we would love to discuss your involvement. Please email us at info@tagds.com for more details on sponsorship information or check out the FAQ section below.
Coming soon!
Where can I find up-to-date news about the conference or changes to the schedule?
We will strive to keep this website up-to-date through the end of the conference, but the for real-time conference updates and news please follow the TAG-DS twitter account (@TAGinDS) for real-time conference updates and news. If you do not have a Twitter account and are having trouble viewing the feed, tools like Twitter Web Viewer will enable you to see our posts without logging in.
How do I submit a paper to the Boston TAG Party?
The OpenReview submission website can be found here: https://openreview.net/group?id=TAG-DS/2026/Conference. More details are available in the 'Tools for Authors' section above.
Does my paper submission have to follow particular formatting requirements?
Yes, all submissions must use the provided LaTeX template for paper submissions. The template provides three formatting options, for the initial submissions, manuscripts from all three tracks should use the submission option (this is already set as the default) which anonymizes the authors to facilitate the reveiw process. If your manuscript is accepted, you will submitted a camera-ready version using either the proceedings option, for accepted full archival (proceedings) papers or the nonproceedings option for accepted non-archival extended abstracts and accepted open-problem/conjecture track papers. If you have questions, please email us at: info@tagds.com. More specific details on the formatting requirements are available within the template.
What is the deadline for paper submission?
The Boston TAG Party is offering three submission tracks, a full archival paper track, a non-archival extended abstract track, and a open-problem/conjecture track. The deadline for all three tracks is the same: June 15, 2026 - Anywhere On Earth (AOE). A countdown to the deadline is available in the 'Important Dates' section above.
What is the difference between the three submission tracks? Which track is right for my manuscript?
The a full archival paper track must include novel work, results, and theory consistent with requirements for publication as completed research with clear conclusions. If you want your paper to be considered for publication in the conference proceedings, this is the track to which you should submit. Submissions should be no more than 8 pages long, not including references, appendices, or supplementary materials, and should be self-contained, independent of appendices or supplementary material.
The non-archival extended abstract track is less formal. Submissions to this track might include work in progress or summaries of work that has been or will be published in alternative venues. Manuscripts submitted to the non-archival extended abstract track will not be published in the conference proceedings, but will still be presented at the conference if accepted. Submissions should be no more than 4 pages long, not including references.
The open-problem/conjecture track is intended as a way for our community to define some of future directions of TAG research. Submissions should describe important open problems within the TAG space that are broad enough to be accessible and of interest to a significant fraction of the Boston TAG Party attendees but remain focused on TAG topics. Submissions to the open problems/conjectures track will not be considered for publication in the conference proceedings, however we are working to create a public repository for these ideas for ongoing community engagement. More details will be available closer to the start of the conference. Submissions should be no more than 3 pages long, not including references.
If you are having trouble deciding which track is right for you, feel free to email us at: info@tagds.com. We would be happy to help.
How do I register to attend the Boston TAG Party?
The registration site is not yet live, but check back soon for more details. We will post a link in the 'Registration' section above when it is available.
What should I do if I want to attend the Boston TAG Party but cannot afford the registration fee?
We do not want money to prevent anyone from being able to participate in the Boston TAG Party. Funds are available to subsidize registration fees. Please email us at info@tagds.com to request support. Check back for more details will once the registration site is live.
I am in Boston and trying to attend the conference, but I can't find the venue. What should I do?
The Boston TAG Part is being held at Northeastern University from August 18-20, 2026 in the Curry Student Center. The physical address is: CSC Ballroom (2nd Floor), Northeastern University, 360 Huntington Ave , Boston, MA 02115. Instructions for public transportation or parking are available in the 'Venue' section above. If you need assistance, please send us an email at: info@tagds.com. We will be checking our emails regularly as the conference draws near.
How do I learn more about the Topological Deep Learning Challenge or submit to the challenge?
The 2026 TDL Challenge is hosted by our friends at TopoBench and GraphUniverse and looks to be very interesting. You can find an overview of the challenge in the 'Topological Deep Learning Challenge' section above, and you can read the complete details of the challenge or submit to the challenge at this website: https://geometric-intelligence.github.io/topobench/tdl-challenge-2026/index.html. Final submissions are due by August 12th, 2026 - Anywhere On Earth (AOE). A countdown to the deadline is available in the 'Important Dates' section above. The results of the challenge will be announced and the awards will be presented at the Boston TAG Party.
My company/organization would like to get involved. How can we become a sponsor for the Boston TAG Party and what are the benefits?
Great! It is wonderful to have people invested in our active and growing TAG-DS research community and we would love to discuss your involvement. Please email us at info@tagds.com. Sponsors at the Bronze Tier ($500) receive their logo on website. Sponsors at the Silver Tier ($1500) receive the opportunity to give a 5 minute presentation at the conference and their logo on website and event-day signage. Sponsors at the Gold Tier ($3000+) receive the opportunity to give a 15 minute presentation at the conference, their logo on website and event-day signage, a seat on an industry/domain-focused panel at the event, and acknowledgement in Opening & Closing Remarks.