This will be the second edition of the NGT Whiteboard Sessions (NGT WHISE), our discussion-driven initiative designed to strengthen collaboration across Work Packages and create space for new ideas at an early stage.
Following the strong interest in the first session, the next NGT WHISE will focus on a timely and fast-moving topic:
Theme: AI Agents
This edition will be hands-on, interactive, and challenge-based.
Guiding question:
Can AI agents act as optimisers of optimisers?
Starting from a real-world machine learning algorithm developed for a physics use case under hardware deployment constraints, participants will work in groups to prompt and steer AI agents towards the best possible solution.
In the spirit of community-driven research, teams will compete to reach the top of the HEP Bench leaderboard on the NGT Challenges Platform. Only solutions developed entirely through AI agent workflows will be eligible.
Access to an open-source AI agent framework and language model will be provided during the session.
As with all WHISE events, the format will remain lightweight and discussion-oriented, with a strong emphasis on experimentation, collaboration, and exchange of ideas.
Audience
This session is particularly aimed at early-career researchers, PhD students, postdocs, fellows, and junior contributors across NGT, with the goal of creating an open environment where new perspectives can be explored freely.
Coffee and snacks will be provided. A social gathering at R1 will follow after the session.
Please register so we know how many participants to expect.