Generally speaking, the similarity index should be no more than 25%. However, similarity is not the only issue that you have to pay attention to. You have to provide new results and give more discussion in the new manuscript. In addition, you have to provide new contributions and novelties rather than extending the former work directly.
Dear colleague, it is possible for a 0% plagiarism papers but accepted level may be determined by the postgraduate school, or journal or even a conference committee.
In my opinion, technical papers should have a similarity index of about 15-20%, while other papers that have less technical characteristics should be around 5-10 % plagiarism.
If this is work you presented at a conference, and not a published conference proceedings, or previously published, then there is no issue, the presentation and feedback received is part of the process of revisioning for eventual publication.
Generally speaking, the similarity index should be no more than 25%. However, similarity is not the only issue that you have to pay attention to. You have to provide new results and give more discussion in the new manuscript. In addition, you have to provide new contributions and novelties rather than extending the former work directly.