OptoRigBench
A simulated all-optical microscope rig for evaluating AI agents: run records, viewer and interactive tools.
Viewer • Updated • 865 • 983 • 1Note Run records of every episode: steps, images and grades. Scored headline and ablations: `rounds/2026-11-08_*` (code 19a5c8b). Experiments E1–E3 and the E5 scenario sweep: `experiments/`. Earlier folders are development history.
OptoRigBench viewer
🚀AI agents operating a simulated neuroscience microscope
Note Step-by-step viewer of all 250 scored episodes of 2026-11-08 (195 headline in three runs, mid1, frontier1 and small1, plus 55 ablations), with the E1–E3 experiments, the E5 sweep, set-aside attempts and development history.
OptoRigBench Tools
🔬Culture bench and replay tool for OptoRigBench
Note Interactive culture bench (practice seeds) and live replay of any scored episode.