Unfortunately, I do not have much time to revise your new report. Some high-level comments:
Add headers to the methodology and results to split the different parts of the paper attention maps, anomaly detection, and disentanglement loss.
Some of the details you should skip (the big tables). You can possibly make a reference to your repo for more details :), for example for Tables 3 and 4, otherwise, your document is too big.
Make the images bigger (Figure 1) possibly you can use fewer images.
Table 2 can possibly be split into two parts and fitted in the same row to gain some space.
Unfortunately, I do not have much time to revise your new report. Some high-level comments:
Good luck with your submission anyway :)