Tau Reconstruction Meeting

Friday 25 Sept 2026, 12:00 → 13:00 Europe/Zurich

Laurits Tani (National Institute of Chemical Physics and Biophysics (EE)), Michel Hernandez Villanueva (Brookhaven National Laboratory (US))

Live Notes

FCC Tau Reconstruction - September 25, 2026

Present: Michel Villanueva, Laurits Tani, Karl Ehataht, Mohadeseh, Maria Cepeda

Benchmarking

• For today, let’s try to define a small list of benchmarks - tau properties, metrics
• Basic properties
  • Charge, decay mode, ID, 4-momentum
  • Decay vertex, information of daughters
• Benchmark metrics
  • TauID, misID rate, confusion matrix for the most common decay modes, purity
  • Karl: Look as well on the resolution for the momentum of the reconstructed tau
• Decays modes, analyses
  • 𝐻→𝜏𝜏
  • 𝜏 lifetime
    • Vertexing
  • SM decay modes, tau polarization
    • Charged, neutral reconstruction
    • Confusion matrix
• List of analyses/people at
  • https://github.com/HEP-FCC/FCCeePhysicsStudies/blob/master/phys-groups/flavour/TauEW/README.md
• Depending of the detector is the PF algorithm that is being used
• Kaon vs pion?
  • Not highest priority, but important as well to extract |Vus|

Datasets

• Q: how frequently we can request MC when the detectors implement modifications?
  • Don’t know, but we can simulate our own MC for signal modes
  • The full simulation will be ~ TB per dataset
• Q: Do we have space allocated at EOS for this activity
  • Good point, we will check with Tau conveners

Round table

• We have matched analyses - people to follow up how reconstruction is performed
  • Contact people and gather information on the reconstruction
  • A survey is an idea
  • Ask about the background
• Which fraction of people has considered using the ML reconstruction
  • Laurits thinks none so far, but a Gaudi algorithm is in preparation and this can bring more interest
  • Maria: people from B factories may not use it (depends on the analysis), but BSM for sure
  • Michel: For tagging tau pairs, it can be extremely useful. Then the signal side can be heuristic reconstruction
• Yet another aspect: Comparison with ML-based PandoraPF
• Sofia made comparisons on tau ML vs heuristic
• Not a benchmark, but at some point we would like to check how well we are doing on 𝑍→𝜏ℓ and 𝐻→𝜏ℓ
• Mohadeseh interested on 𝜏→3𝜇 (we can also explore 𝜏→3𝑒 if leptonID available)

AOB

• Check the mailing list configuration, seems not everyone is getting the notifications

12:00 → 12:10 Introduction

Speakers: Laurits Tani (National Institute of Chemical Physics and Biophysics (EE)), Michel Hernandez Villanueva (Brookhaven National Laboratory (US))

12:10 → 12:40 Round table

Benchmarking reconstruction.pdf