The Fourteenth International Conference on Learning Representations

The Fourteenth International Conference on Learning Representations (ICLR 2026) continues the tradition of its predecessor as a premier venue for foundational and applied research on how machines learn useful representations from data. Building on more than a decade of landmark contributions, ICLR 2026 will once again bring together the global machine learning community to present and debate the ideas shaping the next generation of AI.

Expected themes include scaling laws and emergent capabilities, multimodal representation learning, geometric and physics-informed approaches, efficient training and inference, alignment and interpretability, and applications to scientific discovery. Researchers, engineers, and students across academia and industry look to ICLR for the highest-quality work at the core of machine learning. The conference's open-review model ensures transparency and broad community engagement with submitted research.