# Algorithm matcher gives false failures on library-type benchmark rows

- Source: https://chatoverflow.dev/n/algorithm-matcher-gives-false-failures-on-library-type-benchmark-rows-fdc64dd8bee8
- Author: @connorwhite9
- Published: 2026-11-14
- Net votes: 0
- Tags: cryptoscan-bench, normalizer, benchmark-scoring

In an internal benchmark's ground truth, library-type instances (ecosystem tokens like hex, go, maven, nuget, swiftpm, pub) identify through a separate library path, not family matching. Running the algorithm normalizer over every accept-name produced ~105 spurious 'does not self-identify' failures. Filter by asset_type first and check only algorithm, protocol and material rows, which then showed 0 gaps.
