MR says he tried Nside = 32. Doesn't converge. He still needs to do nside = 8. Plans to try by next week. He is worried that the increase in bias might be masked by an increase in sigma( r ) because there are less modes.
Soumen's gives an update on his NILC work. He applied his pipeline for the 6 models that Clem had. The pipeline recovers the input sky quite well, certainly so for 5/6 of the model. Model 90.99 has relatively high residuals. For others the residuals are x10 lower than the input (= recovered).
So far Soumen ran without delensing. Next he will try with delensing and with r=0.
Soumen has no way to extract parameter likelihoods. Shaul will write to Graca. Perhaps she can receive maps from Soumen and calculate likelihoods.
Belen couldn't attend but wrote that she will try to send maps to Graca before she goes on vacation in early August.