Follow us

Better evaluations for more capable agents

Bay Area Research Gathering

A closed-door research gathering for people working on agent evaluation.

Date
October 14, 2026 · 5:30–8:30 PM PT
Location
South Bay / Palo Alto, CA
Participation
Closed-door · Approval required
Read the summary Request to join · Closed

Summary · October 14, 2026

What we discussed

PhD students, postdocs, and researchers from AI labs, technology companies, and startups met to discuss what agent evaluation needs in order to become a scientific practice. A recurring concern was the distance between benchmark scores and performance in real use.

Participants asked EvalScience to organize working groups with remote talks, provide a regular venue for work in progress, and match mentors with early-career researchers. EvalScience is now drafting a research agenda from these discussions.

About

A focused gathering for the agent evaluation community

We are bringing together people working on agent evaluation to compare how they design, validate, and use evaluations, and to identify where stronger shared methods are needed. Attendance is limited and each registration requires approval.

Taking place shortly after COLM and just before Stanford Reunion Week, the gathering is an opportunity to meet undergraduate researchers, PhD students, professors, alumni, startup builders, and people working in industry who will be in the Bay Area, build the local community, and advance a shared agenda for stronger agent evaluation. The final venue will be shared with guests.

Who we are inviting

Students, professors, alumni, and people building in startups and industry

01

Students & early-career researchers

Undergraduate researchers, PhD students, and postdocs developing research and looking for peers, mentors, or collaborators.

02

Professors & alumni

Faculty and alumni connecting academic work, open research problems, and the wider Bay Area community.

03

Startup & industry builders

Founders, researchers, engineers, and evaluation leads bringing practical problems from agent development and deployment.