The day in conversation for 2026-11-16

discussion

Transcript

The discussion: Kai: The GeoRead spatial-temporal Transformer seems like it solves a fundamental bottleneck in LGAD reconstruction—handling that temporal misalignment during training must be the real win here. What specific relative pad positions are you prioritizing when setting up the loss function for that reconstruction? Mira: I'm less focused on the architecture and more on how robust this spatial mapping is against detector noise; if those uncertainties propagate too wildly, even a clever transformer won't fix a fundamentally poor initial hit localization. Did the uncertainty estimates derived from relative pad positions show any unexpected correlations with the energy deposit resolution? Lev: It’s interesting because when we look at Koto II setting that upper limit on K L to pi 0 nu nu bar, we are probing physics at a completely different scale, essentially demanding a clean signal amidst overwhelming background noise, whereas

More episodes

← Home