PsyDefDetect invites researchers to tackle a novel challenge at the intersection of Clinical Psychology and Natural Language Processing: detecting and classifying psychological defense mechanisms in emotional support dialogues.
Grounded in the clinically validated Defense Mechanism Rating Scales (DMRS) framework, this shared task aims to advance the understanding of unconscious defensive functioning in text.
⚠️ IMPORTANT: All teams must complete the Result Registration Form before April 8, 2026 (AOE). Teams that do not register by this deadline will not be included in the official ranking.
📌 NOTE: The paper submission deadline of April 17, 2026 is firm and cannot be extended, as it is set by the BioNLP workshop. Please plan your writing accordingly.
News
- 2026-04-05 – We will present three shared task awards: Best System Paper, Best Exploration Paper, and Best Interdisciplinary Insight Paper, with certificates awarded on-site at BioNLP@ACL 2026. We encourage all participating teams to submit a system paper. See Paper Submission for details.
- 2026-04-05 – Result registration form is now open. All teams must register by
April 7→ April 8, 2026 (AOE). - 2026-04-05 – Paper submission guidelines are now available, including format and required references.
- 2026-04-04 – End of evaluation period extended to April 7, 2026 (AOE). Submission guidelines and team registration link will be released on April 5.
- 2026-03-15 – Evaluation period begins.
- 2025-12-20 – Task launch on CodaBench with starter baseline kits.
- 2025-12-15 – PsyDefDetect is officially announced as a shared task at the BioNLP@ACL 2026.
FAQ
Where can I download the dataset?
The dataset is hosted on our CodaBench competition page. After registering for the competition, navigate to the Files tab on the left sidebar, then click input_data to download the dataset, which contains the training and test splits.
How do I submit my results?
In the downloaded test.json, add a new field "label" to each entry with your predicted defense level, and save the file as prediction.json. Due to CodaBench's submission rules, you must compress the JSON file into a ZIP archive before uploading. Go to the My Submissions tab and upload the ZIP file. Once scored, you can choose whether to make your result public on the leaderboard.
We also provide baselines to help you get started. Please refer to the Baselines tab on the CodaBench competition page for details.
How do I register my team?
We will release a Google Form for team registration at the end of the evaluation period.
Contact
Please stay up to date by joining our Discord Server and Mailing List. If you have any questions for the organizers, please email us at Hongbin.Na@student.uts.edu.au and cc Zimu.Wang@liverpool.ac.uk.
Organizers
- Hongbin Na, University of Technology Sydney
- Zimu Wang, Xi’an Jiaotong-Liverpool University
- Zhaoming Chen, University of Utah
- Yining Hua, Harvard University
- Rena Gao, The University of Melbourne
- Kailai Yang, The University of Manchester
- Ling Chen, University of Technology Sydney
- Wei Wang, Xi’an Jiaotong-Liverpool University
- Shaoxiong Ji, ELLIS Institute Finland & University of Turku
- John Torous, Harvard University
- Sophia Ananiadou, The University of Manchester
Acknowledgments
We thank Label Studio for providing Academic Program access to its data labeling platform, which supported our annotation work.
Fun Facts
This psychological defenses dataset was annotated by the author while attending ACL 2025 in Vienna, with a brief visit to the Sigmund Freud Museum.