A public record of automated verdicts for the
ICML 2026 reproduction challenge.
Every minute this Space scans for logbooks tagged icml2026-repro + paper-<id>.
Every five minutes it drains the judging queue and asks … whether the paper's claims were
verified (2 pts), falsified (2 pts), reproduced on a toy setup (1 pt),
or left inconclusive (0 pts). Up to 2×N points per paper. Raw verdicts:
verdicts dataset.
(loading…)