This June session will highlight key insights and developments from the International Conference on Learning Representations (ICLR) 2026.
This technical seminar will explore major themes, emerging trends, and exciting new research from ICLR 2026. Whether you work directly in AI research, apply AI in practice, or are simply interested in understanding where the field is headed, we invite you to join the conversation.
Agenda
- ICLR 2026 Highlights and Key Themes - Vector Institute
- Yuntian Deng: NeuralOS: When Models Simulate an Entire Operating System
- Can an operating system be learned by a neural network? NeuralOS simulates entire GUI environments from user interactions, enabling safer and more scalable training and evaluation of AI agents.
- Dongfu Jiang: StructEval: Benchmarking LLMs’ Capabilities to Generate Structural Outputs. How well can LLMs generate structured outputs? StructEval evaluates formats such as JSON, HTML, React, and SVG, revealing persistent challenges in producing outputs that are reliable for real-world software workflows.
- Yu Bo Gao: DPQuant: Efficient and Private Model Training via Dynamic Quantization Scheduling. DPQuant addresses the challenge of combining differential privacy with efficient low-precision training, delivering significant compute savings while preserving model accuracy and privacy guarantees.
Disclaimer: This event is open to Vector Sponsors, FastLane companies, Vector Researchers, and invited health partners only. If a registration is found not to be eligible, it will be asked to provide verification and, if unable to do so, will not be able to attend the event. Please contact learn@vectorinstitute.ai with any questions.