Xinming Tu @TuXinming
Ranked #9 on Backscroll (2026-07-09) — surfaced for substance, not engagement.
We used BenchGuard (COLM, 2026) while building DrugDiscoveryBench. It caught lots of issues...
Agent + Expert >> Agent/Expert
Github:
https://
github.com/XinmingTu/Benc
hGuard
…
DrugDiscoveryBench:
https://
labs.scale.com/leaderboard/dr
ugdiscoverybench
…