Runs are published successfully, but evaluations remain null and do not appear on profile

Hi OpenML team,

I’m seeing an issue where runs are successfully published and have prediction files attached,
but no evaluations are generated afterwards.

I checked runs from the last 2 months, including recent runs visible from the OpenML run search page sorted by date, and this does not look like an isolated single-run problem. (openml.org)

Examples:

• Run 10596264

  • URL: https://www.openml.org/r/10596264
  • task_id: 31
  • predictions file exists
  • evaluations = null

• Run 10596263

  • URL: https://www.openml.org/r/10596263
  • task_id: 32
  • predictions file exists
  • evaluations = null

For these runs:

• the run object exists and is retrievable
• the prediction file is already uploaded
• but no evaluation values (for example AUC / accuracy) are shown afterwards
• the runs also do not seem to properly show up in the expected profile/run views

From the client side, publishing completed successfully.
So this looks less like a failed submission, and more like the server-side evaluation /
indexing step is not being completed.

Could you please check whether the evaluation backend is currently stuck, or whether these runs
need manual reprocessing?

Thanks!

Start with the OpenML run search page and the example run URLs for 10596264 and 10596263, verifying that prediction files exist while evaluations remain null. Trace the server-side evaluation and indexing path for these runs; done means evaluations such as AUC or accuracy are generated and the runs appear correctly in profile and run views.