U&ME Workshop at ECCV 2026 calls for papers on AI unlearning and editing
Open call for unfinished ideas on model unlearning, editing, and safety.
Get AI news that actually matters
One email a day. Zero fluff. Join 10,000+ professionals.
The U&ME (Unlearning and Model Editing) workshop, to be held at ECCV 2026, has issued a call for papers aimed at researchers exploring fast-moving areas like unlearning, model editing, controllability, and AI safety. The organizers—including the Reddit poster—emphasize that the field is evolving rapidly with many open questions. They specifically welcome submissions from students and researchers with works-in-progress, unusual observations, failed directions, or ideas that don't fit neatly into main conference tracks.
Paper topics include unlearning, model stitching and editing, model merging and “MoErging” (Mixture of Experts Merging), compression, efficient domain adaptation, multi-domain/cross-domain U&ME, online/lifelong learning and unlearning, responsible U&ME (robustness, ethics, fairness, resource efficiency, privacy, regulatory compliance), and applications in computer vision. The workshop aims to bring together thinkers deeply engaged with these problems for rich discussions at ECCV 2026.
- Workshop U&ME at ECCV 2026 focuses on unlearning, model editing, safety, and controllability.
- Encourages submissions of unfinished work, failed experiments, and offbeat ideas.
- Topics include MoErging, lifelong unlearning, domain adaptation, and responsible AI in computer vision.
Why It Matters
This workshop signals growing academic focus on making AI models adaptable, safe, and editable after deployment.