1. Refresher:
	The earlier you do origin correction, the better. Now, the clusters are origin corrected rather than jets.

2. Starts with EM-scale jets rather than LCW jets. Why? 
	Zach: Our calo has different response for electron vs. hadrons (non-compensating). The corrections you apply are different for these objects. LCW is Local Cluster Weights gives weights to clusters based on whether it looks like electron/hadron. EM scale jets you need 30-40% correction for 40 GeV. There was bug in LCW in beginning of Run 2, so we did not use it. Jet group was interested in doing Particle Flow jets (tracks associated with clusters).
	Inside tracker, PF is around 10% correction and outside the tracker, it goes down
	What is the fraction of charged vs neutral in pile up? 2 charged pions vs one neutral, so 2/3 is charged 

3. Non-compensating: Comes from neutrinos and slow energy neutrons. Everything starts with high energy pions which give other hadronic particles. Tevatron used Uranium as absorber to compensate, it didn’t quite work. Richard Wigmans → link this

4. Ask Zach about jet god → Put link here
	 Difference in energy is Y axis and X axis is size → Minimum on this plot (0.3 no pile up and 0.4 now) 

5. Pile up:
	Run 1: mu=6-8, No area based?
	Run 2: mu is much higher. Now we use area based (symmetric pile up subtracted based on the jet area)

6. Jet energy MC calibration is almost all of it: Compare MC/data, apply corrections on pT, eta etc

7. Residual pile up correction: Remove residual pile up based on in-time (NPV) and out-of-time (mu). Parameters are obtained from data, MC and they are different.
	If you generated pile up with Herwig, then we can compare. However, we tune pile up to pp.

8. GSC (Section 5.3):
	Was being developed in Run 1 but not added to the nominal jet calibration then. Now this is well calibrated and we understand variables.
	Order? 1,2 energy deposit info where as 3,4 is track based. The track based addition is completely new, you depend on tracker alignment, etc. 
	nsegments took forever. Important because Punch through is 10% for >1 TeV.
	For CMS, this is a worse problem? Nope because their magnet sits in between. In 2011, all punch throughs went into debug stream.
	Q/g initiated jets. First two variables is different for both and so n_track for q vs. g is different by color factor of 9/4. q/g tagging use these variables as they run on fully calibrated jets because uncalibrated can push ntrack in wrong direction.
	Is there a GSC note? Yes, not GSC paper? Buerocracy! 

	Figure 5: You apply each of them sequentially which neglects correlations. Why? Weren’t good at NN back then, may be we will use them in the future

9. Figure 11: All in situ are shown
	Single particle propagation: Done at super high energy just like before.
10. Figure 12 a. What is flavor composition?
	Zach: Definition - How much is q vs. g? 
	If you look at ttbar with 2 jets, then you are more certain you have bs. This is flavor composition. 
	Flavor response is how does a quark jet look in Herwig vs. Pythia?
	Is there additional uncertainty for light jets vs. b jets? Yes
	Unfolding uncertainty: You need to take into account muon and neutrino from b jet that is missing. B jets are like gluons? Depending on the pT, they look like gluons. 
	B-tagging: All of them get at the displaced vertex rather than jet substructure

11. Figure 12b: Bump at 2.5?  LAr calo has different pulse shape, this is particularly true at eta of 2.5. 2017-2018:
	Took 32 sampling for some to get pulse shape. Run 1 used 5 samplings and now we use 4, why? Because more read out.

12. Figure 6 : Bat man ears – come from inter eta distributions

Fun idea: Look at ITk TDR and point out why we do this?