Word2VecSynonymProvider returns maxSynonymsPerTerm + 1 synonyms when the query term is not among its own k+1 HNSW hits (tied vectors or an approximate-search miss), so Word2VecSynonymFilter emits one synonym more than the documented maximum

Finding LFV-862Area: analysis-common / synonym.word2vec (Word2VecSynonymProvider.getSynonyms result cap; Word2VecSynonymFilter)Priority: lowImpact: silent wrong resultsNovelty: new

Finding page for LFV-862

AI-assisted finding, confirmed by automated checks, not yet triaged by a human Lucene committer. This site files nothing upstream.

What it shows

You configure Word2VecSynonymFilter with maxSynonymsPerTerm=k, whose javadoc calls it the 'maximum number of result returned by the synonym search', so you expect at most k synonyms per token. Word2VecSynonymProvider.getSynonyms searches the HNSW graph for k+1 neighbours on the assumption that the query term is always its own nearest neighbour, drops only the query term and never caps the result, so when the term is not among the hits it returns all k+1. With three terms that share one vector and k=1, getSynonyms(c) returns [a, b] and the filter emits 2 SYNONYM tokens; with 4000 distinct clustered vectors and the factory defaults, 4 terms get 6 synonyms because the approximate search with a beam of 6 does not reach the query term itself, although it is its own strict nearest neighbour by a margin of more than 0.03.

The test shows that with three terms that share one vector, Word2VecSynonymProvider.getSynonyms(c, maxSynonymsPerTerm=1, 0.5f) returns two synonyms.

Expected vs actual

Expected: at most 1 synonym (maxSynonymsPerTerm is the "maximum number of result returned by the synonym search"); actual (-ea and -da): [a, b], because the k+1 HNSW search does not return the query term itself and nothing caps the result.

Files

src/org/apache/lucene/analysis/synonym/word2vec/TestWord2VecSynonymProviderMaxSynonyms.java (JUnit test)

Run it

Tested on Lucene main e1faeba08bde (2026-11-02): still reproduces.
Tested on Lucene main 645d4fe14a20 (2026-10-28): reproduced.
Recorded outcome of run 1 on Lucene main 645d4fe14a20: 1 test(s) run, 1 failed.
How to read the outcome: a failing test is the finding reproducing (the failure message states expected against actual).
Seeds: a rerun with the same seed replays the same random choices. Assertions off (-da -dsa -Dtests.asserts=false) is the production-like setting.

# Lucene checkout at the tested commit, built with: ./gradlew --no-daemon assemble   (JDK 25)
LUCENE=/path/to/lucene; V=12.0.0-SNAPSHOT
CP=$LUCENE/lucene/core/build/libs/lucene-core-$V.jar
CP=$CP:$LUCENE/lucene/codecs/build/libs/lucene-codecs-$V.jar
CP=$CP:$LUCENE/lucene/test-framework/build/libs/lucene-test-framework-$V.jar
CP=$CP:$LUCENE/lucene/analysis/common/build/libs/lucene-analysis-common-$V.jar
CP=$CP:randomizedtesting-runner-2.9.2.jar:junit-4.13.2.jar:hamcrest-3.0.jar   # test-framework dependencies (Maven Central)
CP=$CP:junit-platform-console-standalone-6.1.3.jar   # JUnit Platform console launcher (Maven Central)

javac -d classes -cp "$CP" $(find src -name '*.java')

# run 1: assertions on, seed DEADBEEF
java -ea -esa --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.seed=DEADBEEF \
     -cp "classes:$CP" org.junit.platform.console.ConsoleLauncher \
     execute --select-class org.apache.lucene.analysis.synonym.word2vec.TestWord2VecSynonymProviderMaxSynonyms

Gradle alternative: copy the test into the matching module's src/test/ tree of the checkout and run ./gradlew -p lucene/<module> test --tests <class> -Ptests.seed=<seed> (add -Ptests.asserts=false for a run without assertions).

Feedback

Open an issue in lucene-ai-bug-hunt. Please do not open issues or pull requests in apache/lucene about this finding unless a Lucene committer agrees it should be filed.

Snapshot of 2026-11-04. AI-assisted; not yet triaged by a human Lucene committer. Feedback: open an issue in lucene-ai-bug-hunt.