What you’ll learn:
Which use cases for code models and agent-based systems need trajectory evaluation, and why
Which model skills are essential for autonomous behavior and solving long-horizon tasks in code
How trajectory annotation assists in training and evaluating models
How to ensure the safety of models and agent-based systems
When to use synthetic data, when to get experts involved in data annotation, and how hybrid annotation works
You’ll also find out where researchers are focusing their efforts, and what types of breakthrough products we can expect to be developed next.
Speakers
Aleksei Petrov
Founding Engineer
Boris Yangel
Head of AI R&D
Nikita Pavlichenko
Senior ML Engineer