Reproducibility

Press the OPEN button for more details such as links to the Weights & Biases runs of each experiment. Three runs for each experiment.

Legend

• ❓: no run for the particular experiment.
• ✅: The experiment has been reproduced.
• 🤨🤨🤨: The reproducibility runs show greater performance than the reported numbers.
• ⛔️: The reproducibility failed.