WORKSHOP DESCRIPTION
------------------------
SURGeLLM invites submissions on structured understanding, retrieval, generation, and evaluation in the era of LLMs — with structured artifacts (tables, charts, maps, flowcharts, diagrams) treated as first-class citizens.

Contact: surgellm@googlegroups.com.
See Important Dates: https://surgellm.github.io/acl2026/dates/.
We welcome submissions in two categories: Proceedings (Archival) and Non-Proceedings (Non-Archival).

Workshop format
A full-day workshop with invited talks and community interaction.
Invited talks + panel discussion.
Contributed oral presentations and poster session.
Student mentoring lunch session (planned).

KEYNOTE SPEAKERS
------------------------
Lisa Amini
IBM Research, USA
Google Scholar
Leads IBM’s Data & AI Platforms Research; works on infusing generative and agentic AI into enterprise data platforms.

Dan Roth
University of Pennsylvania and Oracle AI, USA
Google Scholar
Eduardo D. Glandt Distinguished Professor at UPenn; Chief AI Scientist at Oracle; Fellow of AAAS/ACM/AAAI/ACL.

Heng Ji
UIUC, USA
Google Scholar
Professor at UIUC; Co-founder/CTO of Medicas AI; Founding Director of AICE/ASKS; ACL Fellow.

Hamid Palangi
Google, UW, USA
Google Scholar
Staff Research Scienstist at Google, Affiliate Associate Professor at UW.


Submission
------------------------
1) Main Submission (Archival)
Accepted papers are published in the ACL Anthology.
Direct submission (OpenReview)
ARR commitment (OpenReview)

2) Non-Proceedings (Non-Archival)
Presentation-only track for cross-presentation of ACL 2026 Findings and other recently accepted work (not published in the workshop proceedings).

We welcome both long (up to 8 pages) and short (up to 4 pages) submissions in ACL format (excluding references and appendices).
At submission time, authors must indicate whether the paper is intended for archival (proceedings) or non-archival presentation.
ARR-reviewed papers can be submitted via the commitment form if they have not been accepted/published elsewhere.
ACL Findings / main-conference papers may request cross-presentation via the non-archival track (subject to capacity).
We welcome a broad spectrum of contributions: position papers, datasets/benchmarks, systems reports, ablations, and negative results (when well-supported).
If you rely on non-public resources (e.g., proprietary data), ensure results are responsibly summarized and compliant with ACL policies.

Main Submission (Archival)
------------------------
Papers accepted in this category will be published in the ACL Anthology.

Submissions may be under review elsewhere at the time of submission.
Upon acceptance, authors must make a binding commitment to publish the paper in exactly one venue.
Archival submissions must present substantially new and original work and comply with ACL policy on overlap/self-plagiarism.
If you plan to submit your work to a future venue, choose the non-archival track instead.
Submit:

Direct submission (OpenReview):https://openreview.net/group?id=aclweb.org/ACL/2026/Workshop/SURGeLLM#tab-your-consoles
ARR commitment (OpenReview):https://openreview.net/group?id=aclweb.org/ACL/2026/Workshop/SURGeLLM_ARR_Commitment

Non-Proceedings (Non-Archival)
------------------------
This category is intended for presentation and discussion only; papers will not appear in the workshop proceedings.

We welcome cross-presentation of ACL 2026 Findings papers (and other recently accepted work), subject to capacity.
Submit:

Non-archival submission form:https://forms.gle/pStJVCU43ew4hqbs7

What we're looking for
------------------------
We welcome a broad spectrum of contributions, including:
New tasks, benchmarks, and evaluation frameworks for structured reasoning/retrieval/generation.
Models and training recipes (prompting, fine-tuning, structure-specialized embeddings, agentic pipelines).
Systems and deployment lessons (DataOps, governance, reliability, throughput/latency, human factors).
Careful ablations, negative results, and reproducibility studies that illuminate failure modes and trade-offs.

Formatting and length
------------------------
We welcome both long (up to 8 pages) and short (up to 4 pages) papers in ACL format (excluding references and appendices).
Submissions should be consistent with ARR submission requirements: https://aclrollingreview.org/cfp#paper-submission-information
If anonymization is required for your track, do not include author names/affiliations in the submitted PDF.

Submission checklist
------------------------
Pick a track: archival (proceedings) vs non-archival (presentation only).
Use ACL format; stay within page limits (excluding references/appendix).
Follow ARR/ACL anonymization rules when applicable.
Include enough experimental detail to support reproducibility (and clearly state any non-public resources).

Topics
------------------------
Methods for tabular understanding and question answering (multi-table reasoning, multi-hop inference, and multimodal integration).
Natural language interfaces to structured data (Text-to-SQL, semantic parsing, schema linking, and data discovery).
Structure-aware retrieval (tables, cells/rows, chart elements, maps, workflows, and code fragments).
Structured generation (text → tables/charts/figures/code) with faithfulness, controllability, and robust evaluation.
Agentic systems for structured data understanding and analysis (time series, graphs, tabular, etc).
Data-centric AI for LLMs on structured data (representation learning, augmentation, robustness, and domain adaptation).
Benchmarking and evaluation (scalability, throughput, contamination, reproducibility, and human-centered assessment).
Applications and governance for LLMs over structured data (DataOps, privacy, fairness, and reliability).

Important dates
------------------------
Direct paper submission deadline: 2026-03-05
Pre-reviewed ARR commitment deadline: 2026-03-24
Notification of acceptance: 2026-04-28
Camera-ready paper due: 2026-05-12
Workshop dates: 2026-07-02 – 2026-07-03

Mentoring and accessibility
------------------------
We plan a student mentoring lunch session and community Q&A (details TBD).
If you have accessibility needs or questions about hybrid participation, email surgellm@googlegroups.com.

Multiple submission & cross-presentation
------------------------
At submission time, authors must clearly indicate whether their submission is intended for archival (proceedings) or non-archival presentation.
Only papers not committed to any other venue are eligible for archival publication in the ACL Anthology.
Papers previously accepted elsewhere are eligible only for non-archival presentation.
Authors intending to submit their work to future conferences/journals should select the non-archival track.
ACL Findings / main-conference papers may request cross-presentation via the non-archival track (subject to capacity).