AI Paper Reading Group (October)

​Join us for the October Harvard in Tech AI Paper Reading Group, led by José Luis Muñoz, Client Engagement Manager at Innodata and moderated by James Nicholson, Lead Software Engineer at the Harvard Chan School of Public Health.

​Note: this event is only open to Harvard alumni

​This month’s session asks a question that matters to anyone building, buying, or working alongside AI: can AI do economically valuable work at expert level? José Luis will walk through two recent benchmarks that test AI on professional tasks, how each one is designed and graded, and what the results do and don’t tell us about using AI in real workflows.

​Papers we’ll discuss:

• ​GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks: https://arxiv.org/abs/2510.04374

• ​APEX-Agents: https://arxiv.org/abs/2601.14242

​Slides:

​https://docs.google.com/presentation/d/1gylrWxpSvQGT-BrE0lQ9ZmbMuAkx7ZZPKdX1pOWXfIk/edit?usp=sharing