In a relative position encoding scheme, a bias is determined by assigning the interaction between a query at position i and a key at position j to a specific bucket. For a certain range of small, non-negative offsets, this assignment uses a direct one-to-one correspondence, where the bucket index is simply the calculated offset i - j. Given a query at position i=7 and a key at position j=3, which bucket index would be assigned?

Calculating Key Position from Bucket Index

In a relative position encoding system where the bucket index b for a small, non-negative offset i - j is determined by the identity function b(i - j) = i - j, it is true that for every unit increase in the offset, the corresponding bucket index also increases by exactly one unit.