Id-sorted posting-list skipping

Determine the measured skip rates and practical effectiveness of exact top-C merge algorithms on id-sorted posting-list layouts for the misi index.

Background

The paper proves that, for rank-major and identifier-unordered posting lists with uniform within-list weights, an exact merge must read every posting in the selected lists. It notes that identifier-sorted layouts could permit skipping algorithms such as those developed for approximate string joins, but their performance in the misi regime—many lists and low score thresholds—has not yet been established.

References

What remains open, in order of theoretical leverage: skipping on id-sorted layouts (measured skip rates pending), learned within-list impacts (would re-enable WAND at a recall price not yet mapped), calibrated sequential stopping, tiled batch execution (bytes amortize; operations do not), and bandwidth-rich substrates --- the GPU measurement above is a first point on that curve, and bank-level processing-in-memory DRAM is a speculative later one: under the same id-block layout the accumulate is a per-bank local histogram and a verification is an in-bank dot product, though no such device has run this workload.

— misi: a Metric Inverted Sample Index  (2608.27422 - Chavez, 27 Aug 2026) in Section 4, “The merge”; Section 5, “Position”