SweetSpotSimilarity.hyperbolicTf returns max instead of about min for frequencies far below xoffset, because its NaN fallback is right only for the Math.pow overflow above the midpoint, so a hyperbolic-tf subclass is not monotone in freq and ranks a document with one occurrence above one with hundreds

Finding LFV-750Area: misc / similarities (SweetSpotSimilarity hyperbolicTf helper)Priority: lowImpact: silent wrong resultsNovelty: new

Finding page for LFV-750

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

What it shows

You subclass SweetSpotSimilarity so that tf() returns hyperbolicTf(freq), as its javadoc suggests, call setHyperbolicTfFactors(0, 2, 10, 400) and search, expecting the documented curve min+(max-min)/2*(tanh((freq-xoffset)*ln(base))+1): a tf near min (0) for frequencies far below the midpoint 400, rising to max (2) above it. Instead hyperbolicTf returns the maximum 2.0 for every frequency from 1 to 91 and drops to 0.0 at 92, so a document that contains the term once outranks an equally long one that contains it 400 times. Far below the midpoint Math.pow(base, xoffset - freq) overflows to Infinity (once xoffset - freq exceeds about 308 at base 10, about 2705 at the default base 1.3), the ratio becomes NaN, and the guard 'return Float.isNaN(result) ? max : result', which is right only for the overflow above the midpoint, returns max.

The test shows that SweetSpotSimilarity.hyperbolicTf returns max, not about min, for a frequency far below the midpoint: with setHyperbolicTfFactors(0, 2, 10, 400), Math.pow(10, 399) overflows, the ratio is NaN and the NaN guard returns max.

Expected vs actual

Expected hyperbolicTf(1) = 0.0 (the javadoc formula min+(max-min)/2*(tanh((x-xoffset)*ln(base))+1)); actual 2.0 (expected:<0.0> but was:<2.0>), with -ea and -da.

Files

• src/org/apache/lucene/misc/TestSweetSpotHyperbolicTfFarBelowMidpoint.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/misc/build/libs/lucene-misc-$V.jar
CP=$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.misc.TestSweetSpotHyperbolicTfFarBelowMidpoint

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.