OPEN

- An important test of the unfolding is a cross check using the MC truth, which is shown in Figure 12 of the AN. This test should be carried out both ways though: the true M_3 distribution generated with Pythia and the migration matrix taken from Herwig and the other way around. The result of this test should be shown for both rapidity bins. Here it would be good to see a pull distribution, where the pull is defined as (sigma_unfolded - sigma_true) / error_unfolding (it should be sufficient to take the diagonal elements from the covariance matrix for this test). Only then it is possible to see if there is a bias from the method.

>>>

- In the AN at lines 661-662 it says that an estimate of the uncertainty on the hadronisation corrections from the modelling of the hadronisation and the underlying event is taken into account. How is this uncertainty calculated? Since you are using Sherpa, you could directly compare the effect of lund-string versus cluster (AHADIC) hadronisation. It would be good to see, maybe even in a plot, what's the difference of the hadronisation corrections using these two hadronisation models. - How large is the effect of the number of partons that you generate at tree level in Sherpa (NJET parameter) on the hadronisation corrections?

On NP corrections: - Fig 36: we observe here some fluctuations from bin to bin that seem not physical. We wonder if the error bars are correct and if it would not be worth to make larger bin or fit several bins together. After all it looks like a correction independent of m_3.

>>> The NP corrections are currently studied in greater detail.

- Your central final result is a_s(Mz)=0.1194+-0.0011(exp)+-0.0016(pdf). What is the central result if you consider JES and Lumi uncertainties as additive? Do you observe any shift in the central value?

- In the experimental error what are the contributions of JEC, unfolding, lumi and NP? Which is the dominant experimental error? If you follow the method of (f) above please plot the parabolas you get each time you remove an uncertainty source.

>>>

On the results: - the choice of the jet parameter size R of 0.7 is not clearly justified. Indeed increasing the value of R will reduce the K factor (taking it larger would probably still reduce it) but the main point isn't it to have small theoretical scale uncertainties ?

>>>

- What are the alpha_s value obtained with anti-kt jets with R=0.5? Are they consistent with the values for R=0.7? How do the uncertainties compare? Besides uncertainties, we think that this is a very import cross-check of the analysis, although we are aware that it can be a lot of work to get these numbers. There is an alpha_s determination based on an inclusive jet cross section measurement from ATLAS (arXiv:1203.5416), where a large discrepancy between the two values of alpha_s obtained from two jet collections (R=0.4 and R=0.6) is observed. We have to make sure that the alpha_s value obtained is independent of the choice of R.

>>> The alpha_s value is not independent of R, since for R=0.5 in the inner rapidity bin, the theory shows > 10% overall deviation from data, compared with < 1% overall deviation for R=0.7.