ICLR 2025, officially known as The Thirteenth International Conference on Learning Representations, was a premier global gathering dedicated to the advancement of deep learning and representation learning. Held from April 24 to April 28, 2025, the conference took place at the Singapore EXPO in Singapore.
As one of the three most influential conferences in artificial intelligence (alongside NeurIPS and ICML), ICLR is renowned for its open peer-review process and for publishing cutting-edge research across all aspects of deep learning. The event brought together a diverse range of participants, including academic and industrial researchers, entrepreneurs, engineers, and students. The conference covered a broad spectrum of topics, including unsupervised and supervised representation learning, reinforcement learning, computer vision, natural language processing, and societal considerations such as fairness and safety in AI.
As one of the premier global gatherings for deep learning, the conference has evolved significantly to manage its exponential growth. With a record 11,603 submissions and over 11,000 attendees, the traditional model of slide-based presentations for every paper became unfeasible. Consequently, the organization now relies heavily on a tiered poster system to accommodate the volume of high-quality research.
Within this framework, the work “TabWak: A Watermark for Tabular Diffusion Models” was accepted as a regular full paper, ensuring it underwent the complete rigorous review process and appears in the proceedings in its entirety. The paper received the distinction of a Spotlight mention, marking it as a standout contribution despite the poster-only presentation format.
Specifically, “TabWak” was presented during Poster Session 4 on the afternoon of Friday, April 25th (3:00 PM – 5:30 PM SGT) in Hall 3 + Hall 2B. This session hosted a wide variety of topics, typical of the conference’s strategy to group diverse research to facilitate broad networking.
