Date
21 December 2026
About the workshop
Making forgetting a first-class capability of AI
Machine unlearning seeks to remove the influence of designated data, concepts, identities or knowledge from trained models while preserving their broader utility
The need spans privacy and data-removal requirements, copyright concerns, model correction, safety, bias mitigation and responsible AI deployment
MULVA brings together researchers and practitioners working across computer vision, natural language processing, multimodal learning, generative AI and foundation models to discuss methods, evaluation, verification and real-world challenges
Workshop at a glance
Format
Half-day
Talks, presentations and discussion
Research scope
Vision · Language · Multimodal AI
Audience
Academic and industry researchers
Research themes
From forgetting algorithms to trustworthy deployment
The workshop welcomes theoretical, methodological, empirical, systems, benchmarking and application-oriented contributions
Foundations & methods
Exact, approximate, certified and parameter-efficient unlearning; formulations and guarantees
Vision, language & multimodal
Unlearning in computer vision, NLP, LLMs, vision-language systems and foundation models
Evaluation & verification
Benchmarks, metrics, attack-based evaluation and forgetting–retention trade-offs
Privacy & responsible AI
Privacy leakage, copyright, safety, fairness and responsible deployment
Join the conversation
Can we really make AI forget?
Explore open challenges in reliable, efficient and verifiable machine unlearning