Reproducibility Agent Workspace

Paper Reproductions

We let an autonomous agent try to reproduce recent open-access papers from their public code and descriptions — then compare the numbers the paper claims with what actually came out of the runs. Each report shows claimed vs reproduced metrics, the environment used, and what blocked the rest. These are independent machine-run attempts, not peer review.