Fix recall computation for beta-filtered search in benchmark

[magdalendobson]

It turns out that when we implemented the fix to recall computation for filtered search in #1069, the fix did not propagate to beta search. This happened because beta search used the same runner as regular knn search in search/knn.rs, so the wrong GroundTruthMode was passed in. This PR fixes that problem by including GroundTruthMode in SearchSteps so it can be passed in earlier in the stack.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 76.00000% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.11%. Comparing base (ee7dbaa) to head (860ccae).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
diskann-benchmark/src/index/benchmarks.rs	66.66%	3 Missing ⚠️
diskann-benchmark/src/index/search/knn.rs	80.00%	3 Missing ⚠️

❌ Your patch status has failed because the patch coverage (76.00%) is below the target coverage (90.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1238      +/-   ##
==========================================
- Coverage   91.01%   90.11%   -0.90%     
==========================================
  Files         501      504       +3     
  Lines       95572    96074     +502     
==========================================
- Hits        86988    86581     -407     
- Misses       8584     9493     +909     

Flag	Coverage Δ
miri	90.11% <76.00%> (-0.90%)	⬇️
unittests	89.78% <76.00%> (-1.20%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
diskann-benchmark/src/index/inmem/spherical.rs	100.00% <ø> (ø)
...nn-benchmark/src/index/streaming/full_precision.rs	93.61% <100.00%> (+0.06%)	⬆️
diskann-benchmark/src/utils/recall.rs	57.14% <ø> (ø)
diskann-benchmark/src/index/benchmarks.rs	70.77% <66.66%> (+4.11%)	⬆️
diskann-benchmark/src/index/search/knn.rs	78.21% <80.00%> (+0.44%)	⬆️

... and 56 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR fixes recall computation for beta-filtered (and other range-groundtruth) search benchmarks by ensuring the correct GroundTruthMode is threaded into the KNN runner earlier in the call stack, avoiding the previous hardcoded mode mismatch.

Changes:

• Add groundtruth_mode: GroundTruthMode to SearchSteps (defaulting to Fixed) and pass it through Knn::search_all.
• Update KNN runner implementations to accept a groundtruth_mode parameter instead of hardcoding Fixed/Flexible.
• Set GroundTruthMode::Flexible at beta-filtered / multihop / inline-filter benchmark call sites that load range groundtruth.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
diskann-benchmark/src/index/search/knn.rs	Threads GroundTruthMode via SearchSteps into the KNN search runner/aggregator.
diskann-benchmark/src/index/inmem/spherical.rs	Marks beta-filtered/multihop/inline-filter in-memory phases as GroundTruthMode::Flexible.
diskann-benchmark/src/index/benchmarks.rs	Marks beta-filtered/multihop/inline-filter benchmark phases as GroundTruthMode::Flexible.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[harsha-simhadri]

Does the change the results reported in the wiki comparing different filtered search algos?

[magdalendobson]

Does the change the results reported in the wiki comparing different filtered search algos?

No, this error doesn't lead to incorrect recall, but rather to throwing an error if any query point has fewer than k groundtruth results. This wasn't the case for the datasets I tested in the Wiki post. I think I found the error originally when working with smaller datasets where small selectivity led to fewer than k valid results for some points.