Potential Redundancy in Margin Generation Due to Overlapping Contexts?

[Tebmer]

Dear Author,

Thanks for your great work! It is an awesome work and very inspiring!

However, In Section 3.3.1 "Margin generation" of the paper, the margin generation process defines each intermediate context as:
$$context_i = \sum_{j=1}^{i} c_j$$, so accumulative?

This means that later contexts (e.g., $$context_5$$) include all previous chunks ($$c_1, c_2, ..., c_5$$). As a result, the margin generated for $$context_5$$ may duplicate information already extracted in $$context_4$$, leading to redundant margin notes across different chunks and repetitive aggregation results in the end?

Looking forward to your reply! :)