about query aware index

Hello, in the code provided, the query aware attention index is concat with vertical-slash index(https://github.com/ByteDance-Seed/FlexPrefill/blob/main/flex_prefill/ops/flex_prefill_attention.py#L1778). This conflicts to the paper's Algorithm 1. In Algorithm 1, vertical-slash index is not used in query aware attention. I wonder this may affect performance of FlexPrefill?

Start by reading flex_prefill/ops/flex_prefill_attention.py around line 1778 and compare the query-aware attention index construction with Algorithm 1 in the paper. Measure whether including the vertical-slash index changes FlexPrefill performance, then determine whether the implementation or the paper needs correction and record the evidence.

Written by the indexing model from the issue text.