[Puppi] Backport of #44050 (Hadronic tau recovery with protection for fromPV==2 charged particles) to 14_0_X

[nurfikri89]

Backport of #44050

Original PR description:

(Making this PR on behalf of @abenecke)

It was spotted by the ParticleNet team that the efficiency to identify hadronic tau candidates for PUPPI jets at low pT is significantly lower than for CHS jets (https://indico.cern.ch/event/1300663/#2-particlenet-ak4-ak8-tagger-r). The main reason for this is that PUPPI treats particles with fromPV1&2 different than CHS. It was found that PUPPI rejects one prong hadronic taus that most of the time has fromPV==2 and pT < 20 GeV (pT > 20 GeV are protected by PUPPI already). This PR is supposed to recover the loss of hadronic tau candidates by keeping charged particles with pT > 4 GeV and fromPV==2 while maintaining the jet energy resolution, eff and purity for jets. A comparison of different options was done in this presentation: https://indico.cern.ch/event/1364503/#5-puppi-tau-tune

PR validation:

passes the usual runTheMatrix test: runTheMatrix.py -l limited -i all --ibeos

[cmsbuild]

A new Pull Request was created by @nurfikri89 for CMSSW_14_0_X.

It involves the following packages:

• CommonTools/PileupAlgos (reconstruction)

@jfernan2, @mandrenguyen, @cmsbuild can you please review it and eventually sign? Thanks.
@ahinzmann, @gkasieczka, @missirol, @hatakeyamak, @jdolen, @rappoccio this is something you requested to watch as well.
@sextonkennedy, @rappoccio, @antoniovilela you are the release manager for this.

cms-bot commands are listed here

• Backported from [Puppi] Hadronic tau recovery with protection for fromPV==2 charged particles #44050

[cmsbuild]

cms-bot internal usage

[mandrenguyen]

please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-1c3c5e/38515/summary.html
COMMIT: 91f44aa
CMSSW: CMSSW_14_0_X_2024-03-31-0000/el8_amd64_gcc12
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-sw/cmssw/44566/38515/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• You potentially removed 97 lines from the logs
• Reco comparison results: 3993 differences found in the comparisons
• DQMHistoTests: Total files compared: 49
• DQMHistoTests: Total histograms compared: 3346148
• DQMHistoTests: Total failures: 5721
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3340405
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 48 files compared)
• Checked 205 log files, 166 edm output root files, 49 DQM output files
• TriggerResults: no differences found

[mandrenguyen]

type tau, jetmet

[antoniovilela]

hold

• Until sign-off from PPD. @cms-sw/ppd-l2

[cmsbuild]

Pull request has been put on hold by @antoniovilela
They need to issue an unhold command to remove the hold state or L1 can unhold it for all

[malbouis]

If I am not mistaken, this PR is part of the efforts from JME, BTV and TAU to have a unified tagger that uses PUPPI jets as input. @nurfikri89 , can you please confirm?
We would like to have it for 14_0 as agreed with JME.

[nurfikri89]

If I am not mistaken, this PR is part of the efforts from JME, BTV and TAU to have a unified tagger that uses PUPPI jets as input. @nurfikri89 , can you please confirm? We would like to have it for 14_0 as agreed with JME.

Yes. This is my understanding from what @abencke and the PUPPI team want to achieved. We hope that with this PR, the PUPPI tune for 2024 datasets and MCs will be in sync with the planned full re-reco 2022 and & 2023.

[antoniovilela]

unhold

[antoniovilela]

+1

[mandrenguyen]

+1
Changes reco in a closed release, but PPD takes responsibility for this ;-)

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_14_0_X IBs (tests are also fine) and once validation in the development release cycle CMSSW_14_1_X is complete. This pull request will be automatically merged.

[antoniovilela]

+1 Changes reco in a closed release, but PPD takes responsibility for this ;-)

precisely