This blocks 4.2.11

As I understand from @Kgraessle , 4.2.11 is unblocked by directly consuming Liftwing output. Though this task will be fixed when wikidiff is fixed, this is no longer a blocker.

Just noting that 4.2.11a was unblocked by directly consuming Liftwing output. This process took a very long time to retrieve all the scores and won't work for re-running the analysis on different wikis. This would still be a blocker for WE4.2.11b Add RevertRisk filters into Special:RecentChanges on larger wikis. Please reach out to me if you have any further questions, thanks!

(Context: I'm supporting fkaelin with this task if further prioritization or changes are needed.)

@Kgraessle thanks for expanding. I'll get in touch with you outside of this task to learn about WE.4.2.11b and the timelines you have in mind for that hypothesis (I don't see it in the internal hypothesis sheet and as a result I don't see what Research has committed there). For now I'm keeping the delivery time for this task to April 30. I'll be in touch shortly.

@Kgraessle, setting aside the status of the hypothesis, I looked into your request in more details with Fabian's help and here is where we are and recommendations for next steps:

• A portion of your request is focused on the dataset of revert risk language agnostic prediction outputs. That is something that Research is maintaining at the moment. The data for this dataset will be refreshed as soon as T388144 is completed. Due date May 16 though @fkaelin is prioritizing to finish it earlier if possible. (risk_observatory.revert_risk_predictions)

• The second part of your request is related to the dataset of revert risk multilingual prediction outputs. Research does not have pipelines built for this dataset. However, this is data for a model that is in Production and the ML or DPE teams may be able to help you with building a dataset from it.

Next steps:

• I'm going to remove Fabian as an assignee of this task. You can follow T388144 and as soon as that task is closed, you'll know that part 1 of your request is met.
• I'm going to remove Research and will add the tag for ML and DPE teams for them to consider the second part of your request and triage.

@Kgraessle I've logged the second part of your request (re: the multilingual RR model) as a request to the ML team on our intake request tracker. Based on our recent conversations, my understanding is that the Moderator Tools team is moving forward with just the language-agnostic RR model for now. As such, the ML team is deprioritizing this request for support with the multilingual model. Please feel free to re-submit this request using this template if you need the datasets for the multilingual model in the future.

In T388453#10808451, @SSalgaonkar-WMF wrote:

@Kgraessle I've logged the second part of your request (re: the multilingual RR model) as a request to the ML team on our intake request tracker. Based on our recent conversations, my understanding is that the Moderator Tools team is moving forward with just the language-agnostic RR model for now. As such, the ML team is deprioritizing this request for support with the multilingual model. Please feel free to re-submit this request using this template if you need the datasets for the multilingual model in the future.

Sounds good, thanks for the update.

@leila thank you for helping to coordinate.

The language agnostic revert risk scores dataset that backs the risk observatory dashboard (hive table risk_observatory.revert_risk_predictions) are now up-to-date and again updated monthly (data available until end of April as of now).

@JAllemandou can you please evaluate the effort?

This has been done by @fkaelin. The dataset is maintained by a research-team pipeline, it should only be radar for us IMO.

FWIW, there is also now a mediawiki.page_revert_risk_prediction_change.v1 stream and Hive event table now. This isn't backfilled, but but new revert risk predictions should be available there ongoing.

This stream is also exposed publicly at https://stream.wikimedia.org/v2/ui/#/?streams=mediawiki.page_revert_risk_prediction_change.v1.