DocInsights 2026 Challenges at EMNLP 2026
US$5,000+ Prize Pool | System Papers | Workshop Presentations
Can AI systems move beyond extracting text and truly understand complex documents?
We invite researchers, students, and practitioners to participate in the DocInsights 2026 Challenges, organized as part of the Workshop on Document Intelligence and Understanding at EMNLP 2026.
Participants may enter either or both tracks:
DocSem: Answer quantitative questions from PDFs while identifying the exact document evidence supporting each answer.
Dr.DocBench: Convert difficult document images into structured Markdown while preserving text, tables, formulas, reading order, and specialized notation.
The challenges offer a combined prize pool of more than US$5,000. Participants may also submit archival or non-archival system papers, and winning and selected teams will be invited to present their work at DocInsights 2026.
Important Dates
DocSem final submission: September 10, 2026
System paper submission: September 15, 2026
Dr.DocBench final submission: October 10, 2026
DocInsights workshop: October 29, 2026, in Budapest
Links
Challenge overview: https://docinsights-workshop.github.io/docinsights-2026/shared-task/
DocSem: https://huggingface.co/datasets/amitbcp/docinsights-2026-shared-task-data
Dr.DocBench: https://drdocbench-challenge.abaka-pages.com/
System papers: https://openreview.net/group?id=EMNLP/2026/Workshop/DocInsights_Shared_Task
If you work on NLP, Document AI, multimodal models, OCR, information retrieval, or grounded reasoning, we would be excited to see what you build.
Please feel free to circulate this announcement among interested colleagues, students, and research groups.