Tom
Lead, Game
Ex-Roblox world models. Indie game developer.
Knows how to make a generated world actually fun to play. Most video models look real, but the things you can do inside them stay extremely limited, his bar is “as good as a real game.”
Project leads & contributors
The bridge.
AgenticVBench was led by an interdisciplinary team from Philo Labs who know two things at once: how creative work actually gets made, and how agentic RL works.
AgenticVBench is a community project, built on the work of 20 industry experts (editors, colorists, post supervisors, AI directors) and shaped by early feedback from frontier labs. We believe artists should be in the community helping build creative superintelligence, not adjacent to it.
Project leads
Snow
Lead, Film
CCA-trained, award-winning film director. Multimodal researcher.
Knows how to teach an AI to direct, what to cut, when to hold, what a director leaves out.
Yi
Lead, Physics
Cambridge-trained physicist. Physics simulation expert.
Knows how to teach AI video models the way the world actually moves, gravity, collisions, the rules a model gets wrong when all it has seen is pixels.
Christine
Lead, Music & Aesthetics
Stanford-trained AI researcher. Landscape photographer.
Knows how to teach AI to see what a photographer sees, framing, light, what's worth pointing the camera at.
Open call · contributors wanted
Help us build the next benchmarks.
v1.0 is out. Three held-out benchmarks come next, Film Generation, Game Creation, Audio-First, and we're looking for video researchers and artists who want to help shape what each one tests, contribute tasks, and co-author the next release.
Open-source by default. Credit on the paper. Come say hi.
Coming up · CVPR 2026
Find us at CVPR this June.
We're giving several talks and hosting researcher dinners and happy hours. Join the Discord or add @christine2wd on WeChat if you'll be there. Come build or chat with us en route towards creative superintelligence.